The Applications of Intelligent Optimization Algorithms in Sentiment Analysis: A Deep and Systematic Study

Sangkeum Lee^1*, Li Ping Yan², Maryam Sanatifar³, Lingyi Feng⁴, Seyedalireza Jazayerifar⁵ and Atieh Nouralishahi^6
*Correspondence: Sangkeum Lee, Environment ICT Research Section, Electronics and Telecommunications Research Institute (ETRI), 218 Gajeong-ro, Yuseong-gu, Daejeon 34129, Korea, Email:

Keywords

Evolutionary algorithms • Entropy • Sentiment analysis • Entropy • Intelligent computing • Particle swarm optimization • Multi-objective optimization • Metaheuristic optimization algorithms

Introduction

Metaheuristic-based optimization algorithms utilize an easy and powerful approach to tackle issues related to optimization [1,2]. With a set of problemindependent techniques and the ability to comprehend higher-level algorithmic frameworks, meta-heuristics can be used to create optimized approaches [3,4]. Metaheuristic can be defined as a method to get a "good enough" solution in "small enough" computing time. Meta-heuristic algorithms give a better trade-off between exploration (global optima) as well as exploitation (local optima), along with solution quality and computing time [5,6]. Meta-heuristic algorithms are more dependable and effective than precise approaches because they can be adjusted to the demands of real-time optimization problems, produce better solutions with less computation time, and are not problem-specific, approximate, or deterministic [7,8]. Swarm intelligence, as a well-known subset of metaheuristics, is a wide range of techniques that work based on artificial intelligence inspired by the cooperative behavior of basic organisms found in the wild. Its idea came from observations of the group behavior of various species, including birds, fish, bees, bacteria, ants, squirrels, and more [9]. Swarm intelligence algorithms are frequently employed in optimization issues. Optimization techniques could be applied to the feature selection problem to create the best possible feature set. Swarm intelligence techniques are employed in feature subset selection to decrease the feature subsets' computational complexity and dimensionality, improving classification accuracy [10]. However, population diversity is quickly lost when a standard swarm intelligence algorithm encounter challenging multi-peak problems, which causes the algorithm to prematurely converge. Dimension entropy is used to address this issue as a measure of population diversity. Entropy is a concept in thermodynamics. In 1948, Shannon put forward the concept [11]. The concept is that one can gauge how uncertain an information system is by calculating the probabilities of numerous separate, random discrete events. In an information system, information entropy is at its least when only one event occurs and at its maximum when the chance of occurring several discrete events is the same [12].

The World Wide Web (www) has recently become a great source of usergenerated content. Twitter, Facebook, Google Plus, and many social media are available where people freely share their thoughts and feelings [13,14]. They do this through messaging or emojis, pictures, and videos. Undoubtedly, any decision-making process may benefit greatly from this constantly expanding subjective knowledge. Intelligent sentiment analysis was established to automate such data processing [15,16]. The systematic detection, extraction, quantification, and analysis of affective states and subjective data is known as intelligent sentiment analysis. It uses biometrics, computational linguistics, natural language processing, and text analysis. It is a method of expressing an individual's feelings or ideas in text form, which may be categorized as negative or positive [17]. Intelligent sentiment analysis points to the wide area of processing natural language, linguistic computation, and text mining to analyze a person's thoughts, feelings, observations, behaviors, decisions, and emotions [18,19]. Intelligent sentiment analysis of social media data is critical in decision-making in various fields. Corporations, for example, need to know how people feel about their products, and governments need to know how people feel about certain decisions [20]. Therefore, sensitivity analysis is a challenge of textual analysis, but certain problems are impossible to equate with conventional text-based analysis.

So, the literature of the last three years clearly has shown an explosion in the field of intelligent optimization algorithms in sentiment analysis from different sources of inspiration, but so far, no regular and systematic study has been presented in this field. Thus, this paper offers a thorough overview of the important intelligent optimization algorithms used in sentiment analysis. A comprehensive publishing trend of meta-heuristic algorithm selection methods has also been presented. The challenges in the articles are examined, and solutions are provided for them. Finally, this study draws crucial conclusions and recognizes research gaps to motivate researchers in this area. The objectives of the research are clearly explained below.

• Offering a Systematic Literature Review (SLR) and investigating the existing strategies for intelligent sentiment analysis using intelligent optimization algorithms;

• Outlining the crucial fields where the research can upgrade the function of intelligent sentiment analysis in intelligent optimization algorithms;

• Detecting obstacles, prosperity criteria, and total notions that intelligent sentiment analysis may encounter utilizing intelligent optimization algorithms;

• Identifying gaps in previous studies and providing solutions for future studies.

• Seven sections are arranged in the remainder of the document. The literature review of intelligent sentiment analysis utilizing intelligent optimization algorithms is viewed in Section 2. The suggested approach for feature selection is explained in Section 3. Section 4 covers the selected papers. Section 5 presents the discussions and results. So, challenges are provided in Section 6. The paper ends in Section 7 by providing its upcoming scope.

Literature Review and Related Work

Clients are gradually making their viewpoints and insights accessible online today [21]. This condition is developing an increasing interest in technology for the automated processing and mining of thoughts and sentiments from web records. The lately updated read-write network has led to a rapid creation of user-written content that results in a vast volume of redundant data. Hence, intelligent sentiment analysis is becoming increasingly critical as the number of digital text resources increases in parallel with the development of information technology [22,23]. Intelligent sentiment analysis has been the one technique used in the last couple of years that processes the available online-unstructured data to obtain important information and recognize the user's sentiments [15]. Intelligent sentiment analysis is research that demonstrates individuals' thoughts, feelings, assessments, appraisals, and feelings against various institutions in the form of a text. Recently, curiosity has grown significantly in recognizing multiple sentiments and elements simultaneously [24]. The study of sentiment is a new area of text-based research. Opinion mining or intelligent sentiment analysis examines methods to address a corporation's issues of public opinion, perceptions, and feelings with which the entity may represent incidents, individuals, or subjects [25]. Intelligence sentiment analysis research can be divided into three primary techniques: Machine Learning (ML)-based, lexicon-based, and hybrid. Some researchers have recently been using hybrid technology for better performance [20].

On the other hand, the development of species in Darwinian natural selection theory, ant behaviors in biology, flock behaviors of some birds, and annealing in metallurgy are only a few examples of the natural occurrences that serve as inspiration for meta-heuristic algorithms [26,27]. Intelligent optimization algorithms research methods inspired by nature efficiently resolve concurrent universal challenges. Also, compared to its earlier possibility, it makes it easier to identify rational universal problems timely with more reliability and precision. Furthermore, meta-heuristic algorithms offer the foundation for a wide range of real-world applications by finding effective solutions to a certain crucial event based on a well-defined set of criteria [28-30]. (Table 1) lists various methods of nature-inspired algorithms discovered by the researchers.

The motivation for the study

In this part, several review articles on the subject of this study have been reviewed to highlight the motivation of this study. Ahmad SR, et al. [31] highlighted comparative studies on feature selection in intelligent sentiment analysis using natural language processing and modern methods such as GA and rough set theory. The study compared the selection of features in text categorization based on conventional and intelligent sentiment analysis techniques. They discovered that by removing characteristics that are redundant and irrelevant, meta-heuristic-based algorithms have the potential to be used in intelligent sentiment analysis research and can provide an ideal subset of features. Their study is not systematic. Kumar, Kumar A, et al. [10] studied the state-of-the-art of various swarm intelligence algorithms presently used for feature subset selection within the sentiment analysis framework. They found that there were only a few swarm algorithms that have been applied in the area, and there were many other algorithms that could be explored; their study provided insight into the various algorithms which can be expounded for improved sentiment analysis Shrivastava H, et al. [32] researched the most recent versions of the various swarm intelligence algorithms currently in use for the sentiment analysis framework's feature subset selection. They discovered that only a few swarm algorithms had been used in the field and that many more algorithms could be investigated. Their research has shed light on the various algorithms that can be used to improve sentiment analysis.

As a simple reference guide for identifying the most effective approaches, Ighazran H, et al. [33] summarized the new distributed strategies for feature selection in intelligent sentiment analysis focused on metaheuristics and evolutionary algorithms. Their summary intended to address a particular issue, specifically in the feature selection process in the intelligent sentiment analysis domain. They analyzed feature selection in intelligent sentiment analysis to recognize their strengths and drawbacks. They observed that meta-heuristic algorithms could theoretically be utilized in intelligent sentiment analysis as feature selection. They recommended metaheuristic algorithms for optimal features picked from consumer feedback. Suddle MK and Bashir M [34] examined metaheuristics-based long short-term memory optimization for intelligent sentiment analysis. For the Long Short-Term Memory (LSTM) architecture optimization, they used GA, PSO, DE, Firefly, and cat swarm optimization. In their studies, they used four benchmark datasets for intelligent sentiment analysis. PSO and DE significantly improved the F-score and accuracy. Using the suggested meta-heuristics, experimental results showed that the ideal configuration found for creating the LSTM architecture considerably increases the accuracy of intelligent sentiment analysis. The most important features of the articles are presented in (Table 2).

Different works developed in the past were focused on intelligent sentiment analysis, metaheuristics, intelligent optimization, and evolutionary algorithms. Over the past few years, sentiment analysis (also called opinion mining, sentiment extraction, intelligent sentiment analysis, opinion extraction, etc.) has experienced a strong rise in scientific research. Investigators have examined several strategies for automating the intelligent sentiment analysis procedure in data mining, natural language processing, Govindarajan M [35]. Despite widespread study in the field, some issues are still unanswered. Besides, the main elements in intelligent sentiment analysis have not been systematically checked out. Besides, as can be seen, most articles have only studied one algorithm, or only one has addressed open issues. The investigator expects the current study to motivate further professional and academic discussions on intelligent sentiment analysis and intelligent optimization algorithms. Available investigations have scarcely checked out these discussions. The present article aims to solve this issue, and paper construction is subordinated to this target. These weaknesses motivated us to write this article to summarize intelligent sentiment analysis using intelligent optimization algorithms comprehensively.

Methodology

This section offers an SLR of intelligent sentiment analysis employing intelligent optimization algorithms in sentiment analysis techniques to convey a clear understanding of intelligent sentiment analysis utilizing intelligent optimization algorithms in sentiment analysis. Firstly, a systematic review is carried out to conduct a complete search of the literature to find relevant research [36]. An SLR is conducted to evaluate the literature and determine relevant studies [37,38]. Some scientific databases are investigated to perform SLR, including Google Scholar, Springer, IEEE, and science direct. The title and the abstract sections of recognized papers were checked for relevance to our issue. Section 3.1 describes the formalization of research questions, and the process of selecting relevant articles is provided in section 3.2.

Question formalization: All reliable and efficient studies that have examined intelligent sentiment analysis using intelligent optimization algorithms are explored in this paper. Furthermore, the prominent features and techniques of each will be considered. Finally, to achieve the purposes mentioned earlier, the selected methods by researchers are recognized, the results are evaluated, and the following study questions are defined:

• RQ1: What are the results of an SLR in intelligent optimization algorithms in sentiment analysis?

• RQ2: What well-liked algorithms are employed to intelligent optimization algorithms in sentiment analysis?

RQ3: What are the next developments, unresolved problems, and difficulties in sentiment analysis using intelligent optimization algorithms?

Article selection method: In the SLR procedure, papers are selected in three stages: automated keyword search, selection of papers based on abstract, title, and publication quality, and analysis of references and publications.

Automated search regarding the main keywords, stage 1: Identifying the journal articles related to intelligent sentiment analysis using intelligent optimization algorithms relying on mechanism and acceptance factors are the main target of the study mechanism. The search process is performed through Google scholar, Springer, IEEE, and science direct, which has paved the way for scholarly literature research. By using alternative and synonym words for the main keywords, the following search strings were determined and used to search inside the database:

• "Intelligent sentiment analysis"

• "Intelligence optimization"

• "Nature-inspired"

• "Meta-heuristic optimization algorithms"

Choosing articles relying on the title, abstract, and publicationquality, stage 2: This stage aims to select the determined experimental screening benchmarks to guarantee that the review includes only high-quality papers and publications. In this way, 18000 publications have been found in stage 1. After finding the articles in this step, we transferred the title of the articles to an Excel file to select the relevant articles. Then, duplicate articles were removed in the first stage (49 duplicate articles). Then the articles that were patents were also excluded from the study (69 articles were excluded). Then the articles that were not available were also removed (882 articles were inaccessible). Also, articles published in non-recognized conferences or nonrecognized journals were excluded from the scope of the research. Therefore, 6000 articles were also removed at this stage. In the next step, the abstract of the articles was done to review the relevant articles. At this stage, 10,000 articles were removed because they were either non-English or had similar titles and were not related to our study.

Final assessment, stage 3: Here, the whole body of the chosen articles from the prior level is investigated to find suitable articles to be reviewed. The papers that have met the following criteria are selected:

• Is published in the intelligent sentiment analysis fields?

• Has it improved some of the intended parameters?

• Has it provided intelligent optimization algorithms?

Therefore, 1000 articles remained, which were reviewed for full-text review. Of these, there were 400 articles whose full text was not available. Of the remaining 600 articles, 283 were review articles excluded from the research scope. Of the remaining 317 articles, 195 were from before 2013 and were excluded from the study. Finally, out of the remaining 122 articles, there were only 31 articles that were really related to the subject of our study, so they were selected for further investigation in this study. (Figure 1) shows the publication of articles in various journals. (Figure 2) illustrates the selected papers' distribution according to the publication year. So, the classification of the selected articles in each group, including Springer, Elsevier, Citeseer, IEEE, MDPI, and IGI, is illustrated in (Figure 2). The authors found 31 articles in this stage, as shown in (Figure 3). So, most of the articles were published in Springer and IEEE journals in 2020.

Analysis of the selected papers: The Internet is now an integral part of daily life. Whether or if an online sales company is good depends on customer feedback. Now, opinions can be posted online not only from family and friends but also from others around the world who may have used things, purchased items online, traveled to locations, or watched movies. They may share their views on the Internet [39]. Intelligent sentiment analysis is the computational study of opinions, attitudes, and emotions toward an entity [40]. The entity might represent individuals, events, or topics. These topics are most likely to be incorporated by reviews. Intelligent sentiment analysis is considered a classification task by employing a classification algorithm or classifier to discover the pattern in text data. The selected articles in the previous section will be analyzed in this section. The full text of all 31 selected articles has been read to classify articles. The existing articles are mostly written about 5 algorithms (CS, GA, PSO, ABC, and ACO). There were also 4 articles in hybrid category. Based on this, the articles into 6 groups have been classified, as shown in (Figure 4).

Genetic algorithm: The GA, invented by John Holland in 1975, is a heuristic search technique intended to imitate the natural evolution phase [41]. A GA operates on a chromosome population, and in terms of fitness, it eventually makes the population healthier [42]. Two chromosomes are chosen for crossover in each version of the algorithm, and they generate two offspring who may have mutations [43]. After this procedure, they will replace chromosomes. The chromosomes, determined from a fitness function, correlate with their fitness values. This method is repeated until the fitness values satisfy specific convergence requirements or a set number of repetitions is reached [44]. Investigators effectively utilize the GA to solve search and optimization issues like traveling salesman issues and job scheduling [45]. In the continuation of this section, the related articles will be reviewed.

Muthia DA [21] suggested that hotel review intelligent sentiment analysis utilizes the naive Bayes algorithm, incorporating data benefit and GA as feature selection strategies. His research turns out text labeling from hotel reviews in the shape of negative or positive. Before and after applying the feature selection process, the calculation was based on naive Bayes' precision. Utilizing 10- fold cross validation, the validation was carried out. The precision test was calculated using the roc curve and confusion matrix. His study's outcome was a 78.50 percent to 83 percent increase in the precision of naive Bayes.

Alahmadi DH and Zeng XJ [46] proposed a strategy focused on their online social networks to provide tips for potential clients. So, utilizing a GA to maximize trust parameters, they examined the mutual morale between an active consumer and his/her peers. Besides, intelligent sentiment analysis was conducted using a numerical ranking system to derive mates' opinions. Then, they integrated these derived confidence and sentiment values into a vector regression algorithm model to create rating estimates. The findings demonstrated that there were good encouraging outcomes for the suggested solution efficiency. The key benefit of their finding was that where there was a lack of overt and clear trust or ratings, they could solve this issue by incorporating evidence from online social networks like implicit trust and sentiment in Twitter micro-reviews.

In the sense of emotion recognition, Ferreira L, et al. [47] proposed a GAbased approach that seeks to decrease the data imbalance. Their technique enabled them to research the influence of its implementation in texts written in Brazilian Portuguese, using a method of identifying emotions. Experiments demonstrated that while utilizing the support vector machine classifier for emotion recognition, balancing the corpus may be an option, particularly in a multi-class configuration.

By developing adaptive sentiment lexicons, Keshavarz H and Abadeh MS [44] enhanced the polarity grouping of sentiments in microblogs. Corporate and lexicon-based methods merged in the suggested process, and lexicons were developed from the text. A new GA was suggested to tackle this optimization question and identify lexicons to categorize text. The tests on six datasets were performed. The outcomes conduct better than some methods in the two datasets in terms of accuracy.

In four datasets, the suggested solution even outperformed in f-measure terms. Utilizing the recommended technique on six datasets, the accuracy was above 80% in all of them, and the f-measure was above 80% in four of them. Shrivastava K and Kumar S [48] looked at the rising prevalence of the Hindi language on the Internet and decided to use it for intelligent sentiment analysis. This study examined the hidden sentiments in movie reviews taken from the review section of Hindi-language online newspapers. Their study presented a deep learning-based technique in which a gated recurrent unit network was coupled with a Hindi word-embedding model to solve the problems. Their study also used a GA to design gated recurrent network topology automatically and choose the best optimum hyper-parameters. Compared to other traditional resource-based and ML techniques, the recommended GA model successfully produced advanced performance outcomes on the Hindi movie review dataset. (Table 3) summarizes the main features of the analyzed articles.

Cuckoo search algorithm: The CS was presented by Yang XS and Deb S [49]. Because of its effectiveness and simplicity, the CS has many benefits in tackling some extremely non-linear optimization issues with some engineering apps, as it can preserve a balance between global and local random walks [50]. The parasitic practice of cuckoo birds depositing their eggs in the nests of their hosts inspired the algorithm. The laying and breeding of cuckoo eggs is the first underlying inspiration for creating a modern optimization algorithm [51]. Cuckoos mostly seek to discover the right space for their eggs to increase their lifespans. Cuckoos that emerge from fertilized eggs establish new communities or societies throughout time. There are some habitats in each society. The ideal environment will be found in all of society, and the rest of the population will migrate there. The whole populations attempt to remain close to the best habitat within the closest radius [50]. The main aim is to integrate a series of binary coordinates with each solution, indicating whether or not a certain attribute refers to the following features category. With the selected features, a classifier is educated, encoded by the importance of the eggs. The consistency of the solution is then determined by measuring each nest [24,49]. The linked papers with the CS algorithm will be discussed in the below part.

Pandey AC, et al. [52] suggested a new meta-heuristics strategy relying on k-means and CS. The recommended technique was utilized to find the best cluster-heads from the emotional contents of the Twitter dataset. On multiple Twitter databases, the feasibility of the suggested approach was checked and compared to DE, enhanced CS, PSO, two methods of n-grams, CS, and gauss-based CS. Empirical findings and statistical review validated that the suggested technique conducts better than the current approaches. Kansal V and Kumar R [15] suggested the artificial neural network-based intelligent sentiment analysis method as a classification algorithm. The applied artificial neural network functioned in two steps, first trained the system according to the existing data. If the training data were special, then the precision of the classification would be high. They proposed using a CS method to improve the obtained features by adding a new healthy function. Due to the data sorting, the CS provided appropriate attribute sets based on objective criteria, such as negative and positive sentiments. The present research showed that using preprocessing, optimization, and classification approaches can enhance the system's accuracy.

Kumar A, et al. [53] proposed the binary CS as a binary version of the CS for the optimal feature selection to analyze sentiments on online textual content. Like the support vector machine, the simple investigated learning methods were first introduced with the standard term frequency, the inverse paper frequency model, and the new optimization model of attributes. The findings were recorded using the benchmark Kaggle dataset containing a list of tweets. Based on performance precision, the observations were measured. According to the experimental study, the recommended implementation of a binary CS to optimize feature selection in intelligent sentiment analysis was superior to the fundamental supervised algorithms based on the inverse text frequency score of the classic word frequency.

Kaur G and Kukana P [54] suggested using an intelligent sentiment analysis system that preprocessed tweets before using the American standard code for tokenization-based information exchange. With the neural network intervention of feed-forward backpropagation to distinguish tweets into a positive, negative, and neutral perspective, the optimized CS features substantially increase the overall classification strength of the recommended task. Therefore, in terms of accuracy, recall, and precision, the planned study was to perform better than the present project.

Ahmed SH, et al. [20] developed a structure that can be utilized relying on their clustering to evaluate social media text data sentiments. The proposed architecture consisted of three major elements: similarity finding, data cleaning, and randomized clustering CS. To maximize precision, a solution that blends the resemblance degree was proposed. They combined the CS's power with the levy flight method to cluster the text data. Their architecture was used to determine the ideal number of clusters to demonstrate a text dataset. They utilized the Niek Sanders tweets dataset to confirm the model. In contrast with the other six algorithms, the suggested model attained better efficiency. Mandal S, et al. [50] recommended a method based on the CS algorithm and intelligent sentiment analysis to summarize the text. The text document was summarized using the CS algorithm with a sentiment ranking. The empirical research utilized a database of benchmarks. The suggested model's output was contrasted with some current and human-generated performance in terms of recall-oriented understudy for the Gisting assessment (rouge) ranking. By measuring accuracy, recall, and f-measure, their summary method demonstrated a better outcome in terms of rouge score relative to other current summary systems.

Mohan I and Moorthi M [24] proposed a novel minimum spanning treedependent CS for an intelligent sentiment analysis relying on the aspect. The findings of the CS-minimum spanning-tree-ada boost revealed a positive predictive value for negative polarity. It outperforms the minimal spanning treerandom forest, the minimum spanning tree-ada boost, and the CS-minimum spanning tree-random forest by around 9.04 percent, 5.83 percent, and 6.24 percent, respectively. It outperforms the minimal spanning tree-random forest, the minimum spanning tree-ada boost, and the CS-minimum tree-random forest by around 3.9 percent, 1.42 percent, and 2.72 percent, respectively. A positive predictive value-neutral in the CS-minimum spanning-tree-ada boost is 6.14 percent, 2.72 percent, and 3.88 percent higher than the minimum spanning-tree-random forest minimum spanning-tree-ada boost and the CSminimum spanning tree-random forest, respectively. (Table 4) summarizes the main features of the analyzed articles.

Particle swarm optimization: PSO algorithm is one of the latest heuristic optimization algorithms proposed by Kennedy J and Eberhart R [55]. It was inspired by social animals' activities, like flocking birds, fish training, and swarming of insects. It uses a series of particles that form a swarm. Every single swarm particle (individually) stands for a potential answer to the optimization issue. The issue's solution space is built as a search space in PSO. A swarm is randomly initialized in the search space of an objective function. Particles fly across hyper-dimensional search space, and their locations are modified, relying on individuals' social-psychological ability to imitate other's performances. The particles of particles are then modified, depending on their own neighbors [43]. So, PSO has a simple and rigorous approach for identifying global optima at enormous sample points [56]. In the continuation of this section, the relevant literature is reviewed. In the article, Xiong W, et al. [57] introduced an appraiser, degree adverbs, negations (and)-a scoring approach that utilized appraisers, negations, degree adverbs, and their combinations for Chinese sentence intelligent sentiment analysis. Furthermore, they employed a PSO method to optimize the parameters of the process's regulations. When the PSO algorithm on the testing corpus decides certain parameters, the efficiency of the sentiment categorization for new sentences and texts will be enhanced. The preliminary findings have demonstrated that new research data sets have been improved.

Kumar Gupta D, et al. [58] proposed a feature selection strategy for attribute-dependent intelligent sentiment analysis. It was based on the PSO principle. They utilized conditional random fields as a learning framework. They executed various characteristics that involve attributes of the lexical, syntactic, and semantic stages. They carried out their studies on the benchmark data sets of several-14 collaborative task databases. The evaluation revealed that their suggested technique performs well for both objectives: aspect term extraction and sentiment categorization. Kurniawati I and Pardede HF [59] applied PSO and information gain to choose features to specify a tweet's sentiment. As a classifier, they utilized a support vector machine. The technology was put into the investigation using data from the West Java Governor's election. Their experiments showed that their proposed method was better than using PSO alone as a feature selector.

Jain A, et al. [56] proposed a PSO, hybridized with a neutrosophic set concept to generate a ternary classifier to analyze the sentiment. The framework is designed to understand the intelligent sentiment analysis of the large text. The results showed the performance improvement of the proposed framework. Also, the proposed method can be validated with other real-time datasets to investigate the accuracy rate and rate of misclassification. Kartiko M [60] analyzed the accuracy of students' sentiments about e-learning using Indonesian on Twitter social media, both positive and negative opinions. Then, the PSO approach was used to optimize the accuracy of the calculation results. In order to optimize accurate results, their study used three experimental sequences (scenario 1, scenario 2, and scenario 2) for both naïve Bayes algorithm and naïve Bayes algorithm-PSO algorithms. The experiment results showed that in scenario 1 an increase in accuracy was 10.00% for the naïve Bayes algorithm PSO. In scenario 2, there was an increase in accuracy of 13.33% on the naïve Bayes algorithm-PSO. Meanwhile, in scenario 3 an increase in accuracy was 27.22% for the naïve Bayes algorithm- PSO. These results proved that the accuracy of naïve Bayes algorithm-PSO was better than naïve Bayes algorithm for all scenarios.

Alfianti ZI, et al. [61] proposed intelligent sentiment analysis of cosmetic review using naive Bayes and support vector machine methods based on PSO. The data examined used data obtained from the website femaledaily. com; the data was taken from a review of cosmetics from four well-known cosmetic brands: Maybelline, Emina, and Wardah. Comparison of accuracy resulting from testing the data was support vector machine algorithm of 89.20% and AUC (area under curve) of 0.973, then with vector machine-based PSO support with an accuracy of 94.60% and AUC of 0.985. The results showed that applying PSO optimization could improve accuracy and provide a more accurate and optimal solution. Intelligent sentiment analysis in Cahyani AD, et al. [62] aims to determine the opinions given by dana and isaku digital wallet service users, whether they contain positive or negative opinions, and apply the naïve Bayes classifier and PSO method to the intelligent sentiment analysis of digital wallet service users. To improve the classification accuracy of the naive Bayes classifier, the research added an attribute weighting method, namely PSO. This study used up to 490 tweets' worth of data from Twitter. The accuracy of the naive Bayes classifiers used in the tests using the confusion matrix and roc curve increased from 53.23% to 85.00% for the i.saku digital wallet and from 60.00% to 91.67% for the dana digital wallet. The results of the T-test and ANOVA tests revealed that the accuracy values for the two classification methods tested differed significantly.

To distinguish thoughts from Twitter accounts, Hayatin N, et al. [63] optimized the naive Bayes algorithm with PSO to achieve a reasonable intelligent sentiment analysis accuracy. PSO was utilized to choose features to discover optimization values for enhancing the accuracy of naïve Bayes. Four levels were required for optimizing the naïve Bayes algorithm utilizing PSO. The tweet group was gained in 2019 on the basis of the community's negative and positive sentiments regarding two Indonesian presidential candidates. The naïve Bayes algorithm- PSO test showed a 90.74% accuracy outcome. The accuracy outcome of the naïve Bayes algorithm was increased by 4.12%. Finally, Machova K, et al. [64] offered empirical outcomes utilizing a natureinspired algorithm—PSO—for labeling. This optimization technique repetitively labels all words in a lexicon and assesses the efficiency of categorizing opinions utilizing the lexicon until the optimal labels have been discovered for words in the dictionary. Also, a framework was integrated into the technique, relying on an ML approach. Their hybrid method categorized more than 99% of texts and accomplished better outcomes than the actual process relying on lexicons. The ultimate hybrid model can be utilized for emotion analysis in human-robot interactions. (Table 5) summarizes the main features of the analyzed articles.

Ant colony optimization: Intelligent is a relatively recent issue-solving method based on insects and other animals [65]. Ants have inspired various approaches and strategies, the most well-known and effective of which is the general-purpose optimization methodology known as ACO. The foraging behavior of various ant species is used to inspire ACO. These ants leave pheromones on the ground to designate a preferred path for other colony members to pursue [66]. Artificial ants, a set of software agents, search ACO for appropriate solutions to a given optimization problem. The optimization issue is converted into the challenge of finding the optimal path on a weighted graph in order to use ACO [67]. ACO is the most foundational swarm-based solution for solving discrete issues [68]. Related articles are reviewed below. Liu X and Fu H [69] presented an ACO-based clustering method to tackle the unsupervised clustering issue using stochastic best solution kept-ESacc. Furthermore, the new algorithm used the Jaccard index to determine the best cluster number. Investigations on three datasets revealed that the algorithm- ESacc takes less time to execute, has a better clustering impact, and was more stable than Sacc. Experimental findings validated the effectiveness of the new method.

Kaur J, et al. [70] looked at Twitter data to see how people felt about a certain topic. Positive and negative opinions were assigned to the tweets that were retrieved. A hybrid ML algorithm, support vector machine, and ACO were used to achieve this categorization. Unigrams were used to extract features, with the feature weighting criteria being term frequency-inverse document frequency. Their work showed that ACO affects the accuracy of support vector machines. The best-achieved accuracy was 86.74% by the support vector machine-ACO model. Ahmad SR, et al. [71] presented a feature selection method that used ACO and k-nearest neighbor algorithms to pick and choose important customer review datasets. In a performance comparison with the suggested method, GA, information gain, and rough set attribute minimization were utilized as baseline algorithms. Precision, recall, and F-score were employed to assess the performance of the ACO-k-nearest neighbor method, which was authenticated utilizing parametric statistical significance tests. This method was statistically demonstrated to be considerably better than the baseline techniques during the assessment phase. Furthermore, the findings of the experiments showed that the ACO-k-nearest neighbor approach might be utilized as a feature selection technique in intelligent sentiment analysis to generate a high-quality, optimum feature subset that accurately represents the real data in customer review data.

Finally, Ahmad SR, et al. [72] suggested a novel strategy for selecting text features that combined a wrapper method with ACO to assist the feature selection procedure. It also evaluated and generated a proposed subset of optimal features using the k-nearest neighbor as a classifier. To identify the link between the feature and the sentiment word in customer reviews, algorithm dependency relations were employed to evaluate the subset of optimal features. The result of the feature subset was utilized as an input to detect and extract sentiment words from sentences in customer reviews, which was produced utilizing the suggested ACO-k-nearest neighbor method. The accuracy of the resultant connection between features and sentiment words was examined and assessed using recall, precision, and F-score. The results revealed that the suggested ACO-k-nearest neighbor technique could obtain the best subset of features and improve sentiment classification accuracy. (Table 6) shows a summary of the articles analyzed in this section.

Artificial bee colony: The ABC algorithm is one of the most newly announced swarm-based algorithms [73]. This algorithm mimics a honeybee swarm's sophisticated foraging activity [74]. It is an optimization method based on honey-bee swarms' unique cognitive behavior [75]. The algorithm is based on the model suggested by Tereshko V and Loengarov A [76] for the honey bee colonies' foraging behavior. Three essential components make up the model: food sources, foraging bees, and employed and unemployed bees. Foraging bees, who make up the first two elements, search for plentiful food sources nearby their hive, which makes up the third element. The model also pinpoints two essential behaviors: forager recruitment to rich food sources, which results in positive feedback, and forager desertion of low food sources, which results in negative feedback—searching for plentiful artificial food supplies in an ABC, a colony of artificial forager bees (agents) (good solutions for a given problem). First, using an ABC transforms the optimization issue to find the optimal parameter vector that minimizes an objective function [77]. The artificial bees then randomly select an initial population of solution vectors and repeatedly improve them by employing the following techniques: use a neighbor search method, moving forward with superior ideas while discarding subpar ones [78]. The following articles related to ABC will be analyzed.

Dhurve MR and Seth M [79] examined the three types of sentiment polarity in sentences at the aspect level of intelligent sentiment analysis. The three scores learned from the data for each term were used to create the scheme. The classification accuracy improved by using optimal feature selection to reduce feature subset size and computational complexity. Their study used an ABC method to optimize the feature subset selection. K-NN, Naïve Bayes, and clustering methods were employed for sentiment-weighted analysis classification. According to the experiments' findings, the ABC method improved the classification accuracy of the classifiers.

With the help of the JAVA platform, Saravanan TM and Tamilarasi A [80] presented an effective opinion mining and classification method that tested on customer and extra review datasets. The suggested research was divided into four phases: (1) data preprocessing, (2) possible feature extraction, (3) opinion extraction and mining, and (4) opinion categorization. Datasets from diverse online publications were preprocessed and presented as part-ofspeech tagged text. An enhanced high adjective count method was used to part-of-speech tagged text to extract possible features. Finally, the review's recall, precision, f-measure, and accuracy values were used to evaluate them. The suggested work received an average of 94.45 percent precision, 80.6825 percent recall, 82.5175 percent f-measure, and 93.05 percent accuracy values.

Orkphol K and Yang W [81] investigated a unique mix of current techniques for improving K-means for a high-dimensional dataset, such as intelligent sentiment analysis in Microblogging. The method of silhouette analysis was also utilized to determine the best K. SentiWordNet rated each group once it clustered into K groups and examined the sentiment polarity of each group. Compared to the random initial state of centroids, it gives superior K-means results but takes longer to compute. Their research showed that integrating several strategies may considerably enhance the K-means outcome when coping with short and noisy communications. To solve the aspect-level sentiment classification problem, Zhang M, et al. [82] employed an attention technique to enhance word embedding and build complicated word vectors for each aspect in a given task. An attention vector based on attention processes was constructed for each particular work element, consisting of two sub-vectors. An ABC algorithm utilizing a support vector machine classifier maximized the attention vector given a sub-dataset connected to an aspect to improve classification accuracy. The results showed that compared to current models, the vector-convolutional neural network model appears to offer significant advantages to increase accuracy. (Table 7) summarizes the main points of the publications examined in the section on ABC.

Hybrid algorithms: Hybrid algorithms use two or more algorithms to solve optimization problems. The hybrid algorithm selects one (based on the data) or alternates between them throughout the procedure. Typically, this is done to integrate desired properties of each so that the entire algorithm is superior to the parts separately [83,84]. In the continuation of this section, the investigation related to hybrid algorithms will be analyzed. Centered on coupling classification approaches utilizing arcing classifier, a new hybrid classification system was proposed, and its efficiency was evaluated by Govindarajan M [35] in terms of accuracy. Utilizing naive bays and GA, a classifier ensemble was constructed. A comparative analysis of the ensemble methodology's efficacy was carried out in the suggested work for sentiment categorization. The ensemble method extended to sentiment classification tasks to successfully combine diverse feature sets and classify algorithms to synthesize a more precise categorization process. Finally, a hybrid framework was suggested to allow optimal utilization of the best performance offered by the individual base classifiers and the composite solution.

The hybrid naive Bayes algorithm-GA showed a higher categorization accuracy percentage than the base classifiers and elevated the testing time because of the decrement in data dimensions. To discover an optimal feature subset, Yousefpour A, et al. [85] utilized a hybrid approach and two metaheuristic algorithms. Two levels were required for the feature selection task: first, utilizing a hybrid filter and wrapper methods to decrease high-dimensional feature space, various feature subsets were gained; second, to obtain an optimal feature subset, local solutions were incorporated utilizing two metaheuristic algorithms. A significant number of comparison experiments on three commonly used datasets in intelligent sentiment analysis revealed that the recommended feature selection method surpasses existing baseline methods in terms of accuracy.

Results and Discussion

The present investigation aims to overview and summarize different intelligent sentiment analysis areas that use nature-inspired meta-heuristic optimization methods. Accordingly, a review of distributed scholar papers emerging in journals through 2022 was obtained to describe essential intelligent sentiment analysis utilizing nature-inspired meta-heuristic optimization methods. Consequently, the studies were rated based on the study goal, publication year, and outcomes. CS, GA, PSO, ABC, ACO, and hybrid algorithms are this research's six primary intelligent sentiment analysis areas. The CS algorithm has seen the greatest recent development momentum out of all the applications. The findings demonstrated that this algorithm could be a useful assessment tool for intelligent sentiment analysis optimization. Intelligent sentiment analysis is one of the leading data mining areas that struggle to detect and analyze sentimental content commonly accessible on social media. The unrivaled growth in the adoption and penetration of social media networks, such as Google Plus, Twitter, Facebook, etc., has modified users' online contact trends in their everyday lives.

Online usage for consumers was formally highly limited to specialized content like news agencies or companies. By making their own content inside a network of peers, they can communicate effortlessly together more collaboratively these days [52]. For example, Twitter is one of the social networks that people utilize in tweets about such subjects. By utilizing clustering-based approaches, these tweets can be analyzed to detect people's opinions and sentiments. Nevertheless, meta-heuristic-based clustering techniques outperform the conventional strategies for intelligent sentiment analysis due to the Twitter datasets' subjective nature. It is noted that the problem with work is that social networking has enhanced the number of internet subscribers enormously. This increasing number of subscribers further raises the number of reports on blogs, forums, and social sites that need a very precise means of reflecting the perspective and mindset behind the shared text [54]. Hence, future works can produce a proper meta-heuristic algorithm in big data technology that could be a feature selection in intelligent sentiment analysis [33,86].

Available research demonstrated that the CS algorithm is more computationally efficient than the impressive PSO [87]. Also, the generic CS algorithm variant generates the subset of the most effective characteristics in categorization. The best search algorithm that is motivated by the breeding behavior of cuckoos is CS. It gives a summary of the nature-inspired algorithm's applications. The CS algorithm is used in multiple areas: image processing, industry, flood forecasting, wireless sensor networks, recognition of speakers, document clustering, health sector, distributed system shortest path, and job scheduling [88,89]. The cuckoo algorithm conducts different algorithms inspired by nature in terms of enhanced performance and less computational time [51]. Furthermore, our findings revealed that, over the last 10 years, nature-inspired optimization algorithms have rapidly progressed and successfully applied to the design and optimization of intelligent sentiment analysis systems. Finally, hybrid systems are still in their infancy as research fields for vehicle routing difficulties. Furthermore, these techniques have a lot of promise for future research.

This study will aid academics because many investigations have been conducted in the previous ten years describing innovative intelligent sentiment analysis in nature-inspired optimization methods. Unfortunately, the use of non-standard language taken from their inspiration area has become common in explaining algorithms, leading to the creation of publications that are commonly hard to read and understand. The obtained explanations are thought to make it simpler for readers to grasp how these algorithms operate quickly without reading the original articles (Table 8). So, metaheuristics practitioners should find it easier to evaluate, understand, and comprehend work that employs the algorithms mentioned earlier. In our viewpoint, the existing studies provide an excellent starting point for any prospective study directions; we have roughly made all of our results and methods independently available. (Table 9) compares the indicators examined in each study. Investigators have attempted to improve categorization accuracy by distributing programs and utilizing algorithms, as seen in the findings in this section. Also, the lack of scalability is one of the most important disadvantages of the reviewed articles. Categorization accuracy is the primary metric for assessing the classifier's performance- the percentage of correctly classified test samples. A classifier's accuracy pertains to a classifier's potential to properly foretell the label of current or previously unseen data [35]. The feature selection technique will potentially enhance the precision of categorization in text mining by lowering the high-dimensional feature space to a low-dimensional feature space, leading to an optimum subset of features accessible [85].

Open issues and future directions: In this section, several challenges and open issues in the field of intelligent sentiment analysis using Intelligent optimization algorithms are discussed. Although the research area is very vibrant, there is still room for improvement. Most swarm intelligence algorithms guide individuals in the direction of the best answer. That aids them in better and perhaps quicker convergence, which isn't always ideal. Other techniques have been offered to address this issue, such as product graphs resembling an arithmetic crossover in evolutionary algorithms. On the other hand, we cannot vouch for the simplicity of all swarm intelligence methods [90]. Also, the Most Valuable Player Algorithm (MVPA) is a very competitive optimization algorithm; it converges rapidly and more successfully than the compared algorithms. The MVPA is a new algorithm in the swarm intelligence field, which has very good results but has not been used in this field. Therefore, further developments and applications of MVPA would be worth investigating in future studies.

Also, it would be crucial to promote additional theoretical investigations into different human representations. Whether representation—binary or real—is superior for a specific family of issues is still up for debate. The scalability of the techniques, on the other hand, appears to be a significant bottleneck, particularly for issues with hundreds or more features. Nonetheless, examining the impact of parallelization may be a useful future research area. Finally, a possible future direction is the creation of novel methods for visualizing results. Naive Bayes classifier is a prominent text categorization ML method since it is so quick, powerful, and successful in several fields. It has a shortage of high sensitivity to an increased number of characteristics [21]. In addition, the help vector machine classifier is also utilized in several methods to classify emotions in the text due to its strong generalization performance and robustness for high-dimensional data. However, most textual companies traditionally exposed to such approaches are naturally imbalanced. Therefore, sensitive to imbalanced data, the support vector machine assigns most texts to the majority class [47]. On the other hand, with the exponential growth of www and the widespread use of online collaboration resources, there is a growing emphasis on automated intelligent sentiment analysis tools that provide a predictive indicator of "positivity" or "negativity" towards views or social statements. The intelligent sentiment analysis and assessment process, however, faces several difficulties. These difficulties become barriers to analyzing the exact sense of emotions and detecting the required emotion polarity.

Identifying and extracting features is the most significant task [33]. Superior classification outcomes must choose the optimal selection of characteristics to assess online textual content sentiment. The optimal choice of features is a computationally challenging task and requires developing innovative strategies to enhance the classifier's efficiency. Also, the ABC algorithm is a novel optimization approach that has proven competitive with existing populationbased algorithms. Nevertheless, the ABC still has a problem with its solution search equation, which is strong at exploration but not so good at exploitation [91]. Let this review serve as a foundation for further research in this field. Researchers can focus on comparing different swarm intelligence algorithms and their variants and prepare in-depth empirical and analytic research. It would also be interesting to look at such algorithms' computational performance and time complexity and compare their efficiency.

Conclusion

Text sentiment detection has attained a high interest and has evolved exponentially in recent years, owing to online feedback's growing digital availability. Because of the rising accessibility and prominence of opinion-rich resources like shopping websites, review portals, blogs, and social media platforms. Numerous academics have been attracted to perform intelligent sentiment analysis. Many studies are covered intelligent sentiment analysis, though there is still a lack of systematic review of intelligent sentiment analysis using intelligent optimization algorithms. For this reason, the present study has concentrated on the intelligent optimization algorithms that can address numerous intelligent sentiment analysis problems and the core technologies related to date until 2022 to discuss the other fields to be studied. Six key algorithms were examined. The current investigation checks the chosen techniques and evaluates the outcomes. Furthermore, unresolved problems have been shown and argued based on an in-depth examination of 31 of the most important articles among the 18000 key papers found throughout the investigation.

Ultimately, it points out the subsequent directions of intelligent optimization algorithms' effects in intelligent sentiment analysis systems that may be useful for more developments. The illustration effectively demonstrates how difficulties in intelligent sentiment analysis have increased worry. The findings indicated that the algorithm presented in this study could provide highly competitive outputs. The literature shows it executes the best algorithms on most test beds. Additionally, findings from real-world case investigations reveal that PSO, CS, can solve real-world intelligent sentiment analysis issues. CS is the most often used method for resolving intelligent sentiment analysis problems. Thus, our study's key advantage is its well-known intelligent optimization algorithms in intelligent sentiment analysis to give some tactics for the future execution of optimization in intelligent sentiment analysis.

In this part, the research questions are answered.

RQ1: What are the results of an SLR in intelligent optimization algorithms in sentiment analysis?

Response to RQ1: According to the section 3 procedure findings, Springer released the most papers, followed by IEEE. Also, the largest number of published articles is related to 2020.

RQ2: What well-liked algorithms are employed to intelligent optimization algorithms in sentiment analysis?

Response to RQ2: The most important algorithms that have been investigated in this field are CS, PSO, ABC, and ACO, that among them, CS was the most important. However, although the genetic algorithm is not among the intelligent optimization algorithms, the intelligent optimization algorithms articles have taken good ideas from them.

RQ3: What are the next developments, unresolved problems, and difficulties in sentiment analysis using intelligent optimization?

Response to RQ3: The most important issue related to the issue of intelligent optimization algorithms in sentiment analysis is the issue of improving accuracy. Researchers in this field are trying to provide frameworks to improve accuracy, and these frameworks are becoming faster and more reliable day by day. In this study, we encountered several limitations. Finding articles using the exact keyword is one of the limitations. A few articles provided the related keywords directly, and the title did not obviously indicate the subject in some articles. Besides, some articles presented an intelligent sentiment analysis using intelligent optimization algorithms that are not selected for various reasons. The patent publications that were inaccessible or written in non-English languages are omitted from our survey. It is suggested that future studies also examine non-English articles published on intelligent sentiment analysis using nature-i

Funding

This work was supported by the Ideological and Political Education Special Project of Yangtze University

Conflicts of Interest

The authors report no conflict of interest.

References

1. Aryafar, Ahmad, Reza Mikaeil, Sina Shaffiee Haghshenas and Sami Shaffiee Haghshenas. "Application of metaheuristic algorithms to optimal clustering of sawing machine vibration." Measurement 124 (2018): 20-31.

Google Scholar, Crossref, Indexed at

2. Kumar, C., T. Dharma Raj, M. Premkumar and T. Dhanesh Raj. "A new stochastic slime mould optimization algorithm for the estimation of solar photovoltaic cell parameters." Optik 223 (2020): 165277.

Google Scholar, Crossref, Indexed at

3. Ajayi, Oluwafemi, Nnamdi Nwulu and Uyikumhe Damisa. "Application of metaheuristic algorithms in DC-optimal power flow."African J Sci Technol Innov 12 (2020): 867-872.

Google Scholar, Crossref, Indexed at

4. Sefati, Seyedsalar and Nima Jafari Navimipour. "A qos-aware service composition mechanism in the internet of things using a hidden-markov-model-based optimization algorithm." IEEE Internet Things J 8 (2021): 15620-15627.

Google Scholar, Crossref, Indexed at

5. Chen, Ping-Shun and Zhi-Yang Zeng. "Developing two heuristic algorithms with metaheuristic algorithms to improve solutions of optimization problems with soft and hard constraints: An application to nurse rostering problems." Appl Soft Comput 93 (2020): 106336.

Google Scholar, Crossref, Indexed at

6. Heidari, Arash and Nima Jafari Navimipour. "A new SLA-aware method for discovering the cloud services using an improved nature-inspired optimization algorithm." Peer J Comput Sci 7 (2021): e539.

Google Scholar, Crossref, Indexed at

7. Khanduja, Neha, and Bharat Bhushan. "Recent advances and application of metaheuristic algorithms: A survey (2014–2020)." Metaheuristic and Evol Comput: Algo Appl (2021): 207-228.

Google Scholar, Crossref, Indexed at

8. Sadrishojaei, Mahyar, Nima Jafari Navimipour, Midia Reshadi and Mehdi Hosseinzadeh. "Clustered routing method in the internet of things using a moth‐flame optimization algorithm." Int J Commun Syst 34 (2021): e4964.

Google Scholar, Crossref, Indexed at

9. Borowska, Bożena. "Learning competitive swarm optimization." Entropy 24 (2022): 283.

Google Scholar, Crossref, Indexed at

10. Kumar, Akshi, Renu Khorwal and Shweta Chaudhary. "A survey on sentiment analysis using swarm intelligence." Indian J Sci Technol 9 (2016): 1-7.

Google Scholar, Crossref, Indexed at

11. Aliaga, J., G. Crespo and A. N. Proto. "Squeezed states and Shannon entropy." Physical Review A 49 (1994): 5146.

Google Scholar, Crossref, Indexed at

12. Kang, Hongwei, Fengfan Bei, Yong Shen and Xingping Sun, et al. "A diversity model based on dimension entropy and its application to swarm intelligence algorithm." Entropy 23 (2021): 397.

Google Scholar, Crossref, Indexed at

13. Chu, Shu-Chuan, Tao Deng and Hong Cheng. "The role of social media advertising in hospitality, tourism and travel: A literature review and research agenda." Int J Contemp Hosp Manag 32 (2020): 3419-3438.

Google Scholar, Crossref, Indexed at

14. Chu, Shu-Chuan and Juran Kim. "The current state of knowledge on electronic word-of-mouth in advertising research." Int J Advert 37 (2018): 1-13.

Google Scholar, Crossref, Indexed at

15. Kansal, Vani and Rakesh Kumar. "A hybrid approach for financial sentiment analysis using artificial intelligence and cuckoo search." ICACCS (2019):523-528.

Google Scholar, Crossref, Indexed at

16. Almuayqil, Saleh Naif, Mamoona Humayun, N. Z. Jhanjhi and Maram Fahaad Almufareh, et al. "Framework for Improved Sentiment Analysis via Random Minority Oversampling for User Tweet Review Classification." Electronics 11 (2022): 3058.

Google Scholar, Crossref, Indexed at

17. Hadna, Muchammad Shiddieqy, Paulus Insap Santosa and Wing Wahyu Winarno. "Studi literatur tentang perbandingan metode untuk proses analisis sentimen di Twitter." Semin Nas Teknol Inf dan Komun (2016): 57-64.

Google Scholar, Indexed at

18. Wardhani, N. K., R. S. Kurniawan, H. E. N. D. R. A. Setiawan and G. R. A. C. E. Gata, et al . "Sentiment analysis article news coordinator minister of maritime affairs using algorithm naive bayes and support vector machine with particle swarm optimization." J Theor Appl Inf Technol 96 (2018): 8365-8378.

Google Scholar, Indexed at

19. Lin, Ying-Lei, Chi-Ju Lai and Ping-Feng Pai. "Using deep learning techniques in forecasting stock markets by hybrid data with multilingual sentiment analysis." Electronics 11 (2022): 3513.

Google Scholar, Crossref, Indexed at

20. Ahmed, Samar H., Khalid Tawfik Wassif and Emad Nabil. "Clustering based sentiment analysis using randomized clustering cuckoo search algorithm." IJCSNS 20 (2020): 159.

Google Scholar

21. Muthia, Dinda Ayu. "Sentiment analysis of hotel review using naive bayes algorithm and integration of information gain and genetic algorithm as feature selection methods." ISSIT 1(2014): 25-30.

Google Scholar, Crossref, Indexed at

22. Dang, Cach N., María N. Moreno-García and Fernando De la Prieta. "Using hybrid deep learning models of sentiment analysis and item genres in recommender systems for streaming services." Electronics 10 (2021): 2459.

Google Scholar, Crossref, Indexed at

23. Liu, Qi, Gengzhong Feng, Weibo Zheng and Jun Tian. "Managing data quality of cooperative information systems: Model and algorithm." Expert Syst Appl 189 (2022): 116074.

Google Scholar, Crossref, Indexed at

24. Mohan, I and M. Moorthi. "Topic flexible aspect based sentiment analysis using minimum spanning tree with Cuckoo search." JAIHC 12 (2021): 7399-7406.

Google Scholar, Crossref, Indexed at

25. Medhat, Walaa, Ahmed Hassan and Hoda Korashy. "Sentiment analysis algorithms and applications: A survey." Ain Shams Eng J 5 (2014): 1093-1113.

Google Scholar, Crossref, Indexed at

26. Tayarani-N, Mohammad-H., Xin Yao and Hongming Xu. "Meta-heuristic algorithms in car engine design: A literature survey."            IEEE Trans Evol Comput 19 (2014): 609-629.

Google Scholar, Crossref, Indexed at

27. Ghadimi, Noradin, Adel Akbarimajd, Hossein Shayeghi and Oveis Abedinia. "Two stage forecast engine with feature selection technique and improved meta-heuristic algorithm for electricity load forecasting." Energy 161 (2018): 130-142.

Google Scholar, Crossref, Indexed at

28. Sheta, Alaa, Hossam Faris, Malik Braik and Seyedali Mirjalili. "Nature-inspired metaheuristics search algorithms for solving the economic load dispatch problem of power system: a comparison study." Applied nature-inspired computing: algorithms and case studies (2020): 199-230.

Google Scholar, Crossref, Indexed at

29. Odili, Julius Beneoluchi, Mohd Nizam Mohmad Kahar and Shahid Anwar. "African buffalo optimization: A swarm-intelligence technique." Procedia Comput Sci 76 (2015): 443-448.

Google Scholar, Crossref, Indexed at

30. Ou, Mingyu, Lei Liu, Yan Liu and Lanling Lan, et al. "Optimal design of composite achromatic wave plate based on the improvement with initial point selection of intelligent algorithm." Optik 225 (2021): 165722.

Google Scholar, Crossref, Indexed at

31. Ahmad, Siti Rohaidah, Azuraliza Abu Bakar and Mohd Ridzwan Yaakub. "Metaheuristic algorithms for feature selection in sentiment analysis." In 2015 Sci Info Conf (2015): 222-226.

Google Scholar, Crossref, Indexed at

32. Shrivastava, Harshal Thakre1 Vaibhav Patel2 Anurag. "Survey of cuckoo search technique for social media content applying sentiment analysis."IJATER (2018): 13-19

Google Scholar, Crossref, Indexed at

33. Ighazran, Hasna, Larbi Alaoui and Tarik Boujiha. "Metaheuristic and evolutionary methods for feature selection in sentiment analysis (a Comparative Study)." ISAECT (2018) 1-6.

Google Scholar, Crossref, Indexed at

34. Suddle, Mehtab Kiran and Maryam Bashir. "Metaheuristics based long short term memory optimization for sentiment analysis."Appl Soft Compu 131 (2022): 109794.

Google Scholar, Crossref, Indexed at

35. Govindarajan, M. "Sentiment analysis of movie reviews using hybrid method of naive bayes and genetic algorithm." Int J Adv Comput Res 3 (2013): 139.

Google Scholar, Crossref, Indexed at

36. Darbandi, Mehdi, Hamza Mohammed Ridha Al-Khafaji, Seyed Hamid Hosseini Nasab and Ahmad Qasim Mohammad AlHamad, et al. "Blockchain systems in embedded internet of things: systematic literature review, challenges analysis, and future direction suggestions." Electronics 11 (2022): 4020.

Google Scholar, Crossref, Indexed at

37. Heidari, Arash, Nima Jafari Navimipour and Mehmet Unal. "Applications of ML/DL in the management of smart cities and societies based on new trends in information technologies: A systematic literature review." Sustain Cities Soc (2022): 104089.

Google Scholar, Crossref, Indexed at

38. Wang, Guanyu, Dong Shang, Guixin Zhang and Shenglin Zhang, et al. "Effects of QingYi decoction on inflammatory markers in patients with acute pancreatitis: A meta-analysis." Phytomedicine 95 (2022): 153738.

Google Scholar, Crossref, Indexed at

39. Ernawati, Siti and Eka Rini Yulia. "Implementation of the Naïve Bayes algorithm with feature selection using genetic algorithm for sentiment review analysis of fashion online companies." CITSM (2018) 1-5.

Google Scholar, Crossref, Indexed at

40. Primartha, Rifkie, Bayu Adhi Tama, Azhary Arliansyah and Kanda Januar Miraswan. "Decision tree combined with PSO-based feature selection for sentiment analysis."JPCS, IOP Publishing (2019): 012018.

Google Scholar, Crossref, Indexed at

41. Holland, John H. Adaptation in natural and artificial systems: An introductory analysis with applications to biology, control, and artificial intelligence. MIT press (1992).

Google Scholar, Crossref, Indexed at

42. Hamian, Melika, Ayda Darvishan, Mehdi Hosseinzadeh and Milad Janghorban Lariche, et al. "A framework to expedite joint energy-reserve payment cost minimization using a custom-designed method based on mixed integer genetic algorithm." Eng Appl Artif Intell 72 (2018): 203-212.

Google Scholar, Crossref, Indexed at

43. Ghosh, Tarun Kumar, Sanjoy Das and Nabin Ghoshal. "Job scheduling in computational grid using a hybrid algorithm based on genetic algorithm and particle swarm optimization." Adv Intell Inf Syst and Appl Math (2020): 873-885.

Google Scholar, Crossref, Indexed at

44. Keshavarz, Hamidreza and Mohammad Saniee Abadeh. "ALGA: Adaptive lexicon learning using genetic algorithm for sentiment analysis of microblogs." Knowl Based Syst 122 (2017): 1-16.

Google Scholar, Crossref, Indexed at

45. Mane, Sunil B and Sharan Vhanale. "Genetic algorithm approach for obstacle avoidance and path optimization of mobile robot." ICCASP (2018) pp. 649-659.

Google Scholar, Crossref, Indexed at

46. Alahmadi, Dimah Hussain and Xiao-Jun Zeng. "Twitter-based recommender system to address cold start: A genetic algorithm based trust modelling and probabilistic sentiment analysis." IEEE 27th International Conference on Tools with Artificial Intelligence (2015): 1045-1052.

Google Scholar, Crossref, Indexed at

47. Ferreira, Lohann, Mariza Dosciatti, Julio Nievola and Emerson Cabrera Paraiso. "Using a genetic algorithm approach to study the impact of imbalanced corpora in sentiment analysis." Twenty-Eighth International Flairs Conference (2015).

Google Scholar, Crossref, Indexed at

48. Shrivastava, Kush and Shishir Kumar. "A sentiment analysis system for the hindi language by integrating gated recurrent unit with genetic algorithm." Int Arab J Inf Technol 17 (2020): 954-964.

Google Scholar, Crossref, Indexed at

49. Yang, Xin-She and Suash Deb. "Cuckoo search via Lévy flights." In 2009 World congress on nature and biologically inspired computing (NaBIC) (2009): 210-214.

Google Scholar, Indexed at

50. Mandal, Shrabanti, Girish Kumar Singh and Anita Pal. "Text summarization technique by sentiment analysis and cuckoo search algorithm." Comput Engineer Tech Proceed (ICCET) (2020): 357-366.

Google Scholar, Crossref, Indexed at

51. Shrivastava, Harshal Thakre and Vaibhav Patel Anurag. "Survey of cuckoo search technique for social media content applying sentiment analysis.

Google Scholar

52.  Pandey, Avinash Chandra, Dharmveer Singh Rajpoot and Mukesh Saraswat. "Twitter sentiment analysis using hybrid cuckoo search method." Inf Process Manag 53 (2017): 764-779.

Google Scholar, Crossref, Indexed at

53. Kumar, Akshi, Arunima Jaiswal, Shikhar Garg and Shobhit Verma, et al. "Sentiment analysis using cuckoo search for optimized feature selection on Kaggle tweets." Int J Inf Retr Res 9 (2019): 1-15.

Google Scholar, Crossref, Indexed at

54. Kaur, Gurpreet and Poonam Kukana. "Sentiment analysis using cuckoo search and computational intelligence." Int Conf Smart Electron Comm (ICOSEC) (2020): 497-503.

Google Scholar, Crossref, Indexed at

55. Kennedy, James and Russell Eberhart. "Particle swarm optimization." In proceedings of ICNN'95-international conference on neural networks 4 (1995): 1942-1948.

Google Scholar, Indexed at

56. Jain, Amita, Basanti Pal Nandi, Charu Gupta and Devendra KumarTayal. "A hybrid framework based on PSO and neutrosophic set for document level sentiment analysis." Rec Adv Int Inf Syst Appl Math (2020): 372-379.

Google Scholar, Crossref, Indexed at

57. Xiong, Wen, Yaohong Jin and Zhiying Liu. "Chinese Sentiment Analysis Using Appraiser-Degree-Negation Combinations and PSO." J Comput 9 (2014): 1410-1417.

Google Scholar, Crossref, Indexed at

58. Kumar Gupta, Deepak, Kandula Srikanth Reddy and Asif Ekbal. "Pso-asent: Feature selection using particle swarm optimization for aspect based sentiment analysis." Int Conf Appl of Nat Lang Inf Syst (2015): 220-233.

Google Scholar, Crossref, Indexed at

59. Kurniawati, Ika and Hilman F. Pardede. "Hybrid method of information gain and particle swarm optimization for selection of features of SVM-based sentiment analysis." Int Conf Inf Tech Syst Innov (ICITSI) (2018):1-5.

Google Scholar, Crossref, Indexed at

60.  Kartiko, Mohamad. "Accuracy for Sentiment Analysis of Twitter Students on ELearning in Indonesia using Naive Bayes Algorithm Based on Particle Swarm Optimization." J Phy Conf Ser (2019): 012027.

Google Scholar, Crossref, Indexed at

61. Alfianti, Zulia Imami, Deni Gunawan and Ahmad Fikri Amin. "Sentiment analysis of cosmetic review using naive bayes and support vector machine method based on particle swarm optimization." Jurnal Riset Informatika 2 (2020): 169-178.

Google Scholar, Crossref, Indexed at

62. Cahyani, Alvie Delia, Tati Mardiana and Laela Kurniawati. "Sentiment analysis of digital wallet service users using naïve bayes classifier and particle swarm optimization." Jurnal Riset Informatika 2 (2020): 241-250.

Google Scholar, Crossref, Indexed at

63.  Hayatin, Nur, Gita Indah Marthasari and Lia Nuarini. "Optimization of sentiment analysis for Indonesian presidential election using naïve bayes and particle swarm optimization." Jurnal Online Informatika 5 (2020): 81-88.

Google Scholar, Crossref, Indexed at

64. Machova, Kristína, Martin Mikula, Xiaoying Gao and Marian Mach. "Lexicon-based sentiment analysis using the particle swarm optimization." Electron 9 (2020): 1317.

Google Scholar, Crossref, Indexed at

65. Hussein, Mohamed K. and Mohamed H. Mousa. "Efficient task offloading for IoT-based applications in fog computing using ant colony optimization." IEEE Access 8 (2020): 37191-37201.

Google Scholar, Crossref, Indexed at

66.  Dorigo, Marco, Mauro Birattari and Thomas Stutzle. "Ant colony optimization." IEEE Comput Int Maga 1 (2006): 28-39.

Google Scholar, Indexed at

67. Dorigo, Marco. "Ant colony optimization." Scholarpedia 2 (2007): 1461.

Google Scholar, Indexed at

68. Zhao, Dong, Lei Liu, Fanhua Yu and Ali Asghar Heidari, et al. "Ant colony optimization with horizontal and vertical crossover search: Fundamental visions for multi-threshold image segmentation." Expert Syst Appl 167 (2021): 114122.

Google Scholar, Crossref, Indexed at

69. Liu, Xiaoyong and Hui Fu. "An Effective Clustering Algorithm with Ant Colony." J Comput 5 (2010): 598-605.

Google Scholar, Crossref, Indexed at

70. Kaur, Jasleen, Sukhjit Singh Sehra and Sumeet Kaur Sehra. "Sentiment analysis of twitter data using hybrid method of support vector machine and ant colony optimization." Int J Comput Sci Inf Technol Secur (IJCSIS) 14 (2016).

Google Scholar, Indexed at

71. Ahmad, Siti Rohaidah, Nurhafizah Moziyana Mohd Yusop, Azuraliza Abu Bakar and Mohd Ridzwan Yaakub. "Statistical analysis for validating ACO-KNN algorithm as feature selection in sentiment analysis." In AIP conference proceedings 1891 (2017).

Google Scholar, Crossref, Indexed at

72. Ahmad, Siti Rohaidah, Azuraliza Abu Bakar and Mohd Ridzwan Yaakub. "Ant colony optimization for text feature selection in sentiment analysis." Intell Data Anal 23 (2019): 133-158.

Google Scholar, Crossref, Indexed at

73. Ewees, Ahmed A., Mohamed Abd Elaziz, Mohammed AA Al-Qaness and Hassan A. Khalil, et al. "Improved artificial bee colony using sine-cosine algorithm for multi-level thresholding image segmentation." Access 8 (2020): 26304-26315.

Google Scholar, Crossref, Indexed at

74. Karaboga, Dervis and Bahriye Akay. "A comparative study of artificial bee colony algorithm." Appl Math Comput 214 (2009): 108-132.

Google Scholar, Indexed at

75. Karaboga, Dervis and Bahriye Basturk. "On the performance of artificial bee colony algorithm." Appl Soft Comput 8(2008): 687-697.

Google Scholar, Crossref, Indexed at

76. Tereshko, Valery and Andreas Loengarov. "Collective decision making in honey-bee foraging dynamics." Comput Sci Inf Syst 9 (2005): 1.

Google Scholar, Indexed at

77. Chen, Jia, Weiyu Yu, Jing Tian and Li Chen, et al. "Image contrast enhancement using an artificial bee colony algorithm." Swarm Evol Comput 38 (2018): 287-294.

Google Scholar, Crossref, Indexed at

78. Karaboga, Dervis. "Artificial bee colony algorithm."scholarpedia5 (2010): 6915.

Google Scholar, Crossref, Indexed at

79. Dhurve, Ms Ruby and Megha Seth. "A survey on weighted sentiment analysis using artificial bee colony algorithm." (IJARCET) 4 (2015).

Google Scholar, Crossref, Indexed at

80. Saravanan, T. M and A. Tamilarasi. "Effective sentiment analysis for opinion mining using artificial bee colony optimization." Res J Appl Sci 12 (2016): 828-840.

Google Scholar, Crossref, Indexed at

81. Orkphol, Korawit and Wu Yang. "Sentiment analysis on microblogging with K-means clustering and artificial bee colony." Int J Comput Intell 18 (2019): 1950017.

Google Scholar, Crossref, Indexed at

82. Zhang, Ming, Vasile Palade, Yan Wang and Zhicheng Ji. "Attention-based word embeddings using artificial bee colony algorithm for aspect-level sentiment classification." Info Sci 545 (2021): 713-738.

Google Scholar, Crossref, Indexed at

83. Ghadimi, Noradin. "A new hybrid algorithm based on optimal fuzzy controller in multimachine power system." Complexity 21 (2015): 78-93.

Google Scholar, Cross Ref, Indexed at

84. Zhong, Xunyu, Jun Tian, Huosheng Hu and Xiafu Peng. "Hybrid path planning based on safe A* algorithm and adaptive window approach for mobile robot in large-scale dynamic environment." J Int Robot Systems 99 (2020): 65-77.

Google Scholar, Crossref, Indexed at

85. Yousefpour, Alireza, Roliana Ibrahim, Haza Nuzly Abdul Hamed and Takeru Yokoi. "Integrated feature selection methods using metaheuristic algorithms for sentiment analysis." Asian Conf on Int Inf Data Syst 8 (2016): 129-140.

Google Scholar, Crossref, Indexed at

86. Razmjooy, Navid, Fatima Rashid Sheykhahmad and Noradin Ghadimi. "A hybrid neural network–world cup optimization algorithm for melanoma detection." Open Med 13(2018): 9-16.

Google Scholar, Crossref, Indexed at

87. Adnan, Md Akhtaruzzaman and M. A. Razzaque. "A comparative study of particle swarm optimization and Cuckoo search techniques through problem-specific distance function." Int Conf of Inf and Comm Tech (ICoICT) (2013) 88-92.

Google Scholar, Crossref, Indexed at

88. Butpheng, Chanapha, Kuo-Hui Yeh and Hu Xiong. "Security and privacy in IoT-cloud-based e-health systems—A comprehensive review." Sym 12 (2020): 1191.

Google Scholar, Indexed at

89. Xiong, Hu, Chuanjie Jin, Mamoun Alazab and Kuo-Hui Yeh, et al. "On the design of blockchain-based ECDSA with fault-tolerant batch verification protocol for blockchain-enabled IoMT." IEEE j biomed and health inf 26 (2021): 1977-1986.

Google Scholar, Crossref, Indexed at

90. Brezocnik, Lucija, Iztok Fister Jr and Vili Podgorelec. "Swarm intelligence algorithms for feature selection: A review." Appl Sci 8 (2018): 1521.

Google Scholar, Crossref, Indexed at

91. Akay, Bahriye and Dervis Karaboga. "A modified artificial bee colony algorithm for real-parameter optimization." Inf Sci 192 (2012): 120-142.

Google Scholar, Crossref, Indexed at