Abstract Multi-label text classification (MLTC) is a popular method for organizing electronic documents, which is crucial for accessing and processing data. As the number of classes increases, learning multi-label data will be challenging. The number of possible states for various labels increases exponentially, and learning algorithms in single-label data cannot be used to solve these problems. In the meantime, using single-label data algorithms could be very time-consuming. In MLTC, complexity costs should be reduced. Deep-learning neural networks that can learn intricate patterns are used in many real-world problems because of their high power and accuracy. This paper proposed a hybridization of the long short-term memory (LSTM) neural network and the convolutional neural network (CNN) method for MLTC. The proposed model uses LSTM to enhance CNN to improve the proposed model’s accuracy. Also, the competitive search algorithm (CSA) is used to improve the LSTM hyperparameters. The LSTM hyperparameters play an important role in increasing the detection accuracy. The CSA algorithm finds the best values for the hyperparameters by searching the problem space. It was tested on four different datasets of multi-label texts: Reuters-21578, RCV1-v2, EUR-Lex, and Bookmarks. The result showed that the proposed model performed better than CNN and LSTM-CSA in terms of accuracy percentage and that it has improved by an average of more than 10%. Also, the results show that the LSTM-CSA model has higher detection accuracy compared to LSTM—Gradient-based optimizer (GBO) and LSTM—whale optimization algorithm (WOA).

Abstract One of the most important and costly stages in software development is maintenance. Understanding the structure of software will make it easier to maintain it more efficiently. Clustering software modules is thought to be an effective reverse engineering technique for deriving structural models of software from source code. In software module clustering, the most essential objectives are to minimize connections between produced clusters, maximize internal connections within created clusters, and maximize clustering quality. Finding the appropriate software system clustering model is considered an NP‐complete task. The previously proposed approaches' key limitations are their low success rate, low stability, and poor modularization quality. In this paper, for optimal clustering of software modules, Chaotic based heuristic method using a forest optimization algorithm is proposed. The impact of chaos theory on the performance of the other SFLA‐GA and PSO‐GA has also been investigated. The results show that using the logistic chaos approach improves the performance of these methods in the software‐module clustering problem. The performance of chaotic based FOA, SFLA‐GA and PSO‐GA is superior to the other heuristic methods in terms of modularization quality and stability of the results.

Abstract Artificial bee colony (ABC) optimization and imperialist competitive algorithm (ICA) are two famous metaheuristic methods. In ABC, exploration is good because each bee moves toward random neighbors in the first and second phases. In ABC, exploitation is poor because it does not try to examine a promising region of search space carefully to see if it contains a good local minimum. In this study, ICA is considered to improve ABC exploitation, and two novel swarm-based hybrid methods called ABC–ICA and ABC–ICA1 are proposed, which combine the characteristics of ABC and ICA. The proposed methods improve the evaluations results in both continuous and discrete environments compared to the baseline methods. The second method improves the first optimization method as well. Feature selection can be considered to be an optimization problem because selecting the appropriate feature subset is very important and the action of appropriate feature selection has a great influence on the efficiency of classifier algorithms in supervised methods. Therefore, to focus on feature selection is a key issue and is very important. In this study, different discrete versions of the proposed methods have been introduced that can be used in feature selection and feature scoring problems, which have been successful in evaluations. In this study, a problem called cold start is introduced, and a solution is presented that has a great impact on the efficiency of the proposed methods in feature scoring problem. A total of 16 UCI data sets and 2 Amazon data sets have been used for the evaluation of the proposed methods in feature selection problem. The parameters that have been compared are classification accuracy and the number of features required for classification. Also, the proposed methods can be used to create a proper sentiment dictionary. Evaluation results confirm the better performance of the proposed methods in most experiments.

Summary In a vehicular ad‐hoc network (VANET), each vehicle is equipped with an on‐board unit to communicate vehicle to vehicle or vehicle to fixed infrastructure. VANET technology is offered to provide many facilities to passengers and drivers, including safety, entertainment, mobile commerce, driver assistance, and emergency alarms. VANET has unique features such as high‐speed node mobility and network topology dynamics. These special features cause many problems such as increased transmission delays and packet loss. On the other hand, providing a good routing plan for VANET is a critical issue. Therefore, this article proposes a cluster‐based routing using in‐vehicle meta‐heuristic algorithms (CRMHA‐VANET) which has two phases. In the first stage, the vehicles are clustered and the most suitable cluster head (CH) is selected using the gray wolf optimization algorithm (GWO). In the next step, the next suitable CH is selected for data transmission in direct paths using the technique for order of preference by similarity to ideal solution (TOPSIS). The performance of the proposed method is analyzed through several criteria such as package delivery rate, end‐to‐end delay and throughput. CRMHA‐VANET results in a 10% to 25% improvement over all performance metrics, that is, packet delivery rate, latency, and throughput, over CRBP (clustering routing based on PSO [particle swarm optimization]), WCV (weight based clustering for VANET), and AODV‐CD methods.

Abstract Feature selection (FS) in data mining is one of the most challenging and most important activities in pattern recognition. In this article, a new hybrid model of whale optimization algorithm (WOA) and flower pollination algorithm (FPA) is presented for the problem of FS based on the concept of opposition‐based learning (OBL) which name is HWOAFPA. The procedure is that the WOA is run first and at the same time during the run, the WOA population is changed by the OBL. And, to increase the accuracy and speed of convergence, it is used as the initial population of FPA. To evaluate the performance of the proposed method, experiments were carried out in two steps. The experiments were performed on 10 datasets from the UCI data repository and Email spam detection datasets. The results obtained from the first step showed that the proposed method was more successful in terms of the average size of selection and classification accuracy than other basic metaheuristic algorithms. In addition, the results from the second step showed that the proposed method which was a run on the Email spam dataset performed much more accurately than other similar algorithms in terms of accuracy of Email spam detection.

Abstract Solving constrained engineering optimization problems is a highly significant issue, and many different approaches have been proposed in this regard. In this article, a modified farmland fertility algorithm (FFA) has been proposed. This algorithm improves new solutions by benefiting from neighborhoods produced by the new method. In the proposed algorithm, the FFA algorithm phases are modified in the update functions, and some of its variables are replaced with new ones. In the first phase of the algorithm, a mutation is also used to improve the solution of the problem with a particular rule. The results on CEC2019 standard functions were examined to determine the impact of the new parameters. Experiments were performed on 26 standard functions and two constrained engineering optimization problems. To make a better comparison, the analysis of variance, without parameters such as the Friedman test and Pairwise test, was used to evaluate and compare the algorithms. These experiments showed that the proposed algorithm is capable of high‐speed convergence and can minimize most standard functions and engineering constraints within the minimum time and the least function evaluation to the optimum value. The proposed FFA could outperform the previous version of the FFA with minimum modification in convergence, CPU‐time, and complexity.

Abstract Different nature‐inspired meta‐heuristic algorithms have been proposed to solve optimization problems. One of these algorithms is called social spider optimization (SSO) algorithm. Spiders' natural behaviors have inspired them to find the bait position by detecting vibrations in their web. Although the SSO algorithm has good accuracy in achieving optimal solutions, it suffers from a low convergence rate. In this paper, we attempted to improve SSO by changing its motion and mating parameters. To provide a practical example of using the new proposed algorithm, we based it on multi‐objective opposition‐based SSO, named MOPSSO. We used this algorithm in a feature selection process for analyzing text psychology, which is a multi‐objective problem. Textual psychology analysis is used in various fields, including collecting and analyzing people's views on various products, topics, social and political events. After selecting features, in order to classify the text, we used a new hybrid method that hybrids fuzzy C‐MEANS data clustering technique, a decision tree (DT), and Naïve Bayes (NB). Experimental results show that the improved SSO algorithm performs better than SSO, social spider algorithm, and CMA‐ES algorithms. Additionally, the performance of the proposed hybrid classification method is better than those of NB and DT.

Abstract One of the useful knowledge representation tools, which can describe the joint probability distribution between some random variables with a graphical model and can be trained by a dataset, is the Bayesian network (BN). A BN is composed of a network structure and a conditional probability distribution table for each node. Discovering an optimal BN structure is an NP‐hard optimization problem that various meta‐heuristic algorithms are applied to solve this problem by researchers. The genetic algorithms, ant colony optimization, evolutionary programming, artificial bee colony, and bacterial foraging optimization are some of the meta‐heuristic methods to solve this problem using a dataset. Most of these methods are applying a scoring metric to generate the best network structure from a set of candidates. A Fixed Structure Learning Automata‐Based (FSLA‐B) algorithm is presented in this paper to solve the structure learning problem of BNs. There is a fixed structure learning automaton for each pair of vertices in the BN's graph structure in the proposed algorithm. The action of this automaton determines the presence and direction of an edge between the vertices. The proposed algorithm performs a guided search procedure using the FSLA and escapes from local optimums. Several datasets are utilised in this paper to evaluate the performance of the proposed algorithm. By performing various experiments, multiple meta‐heuristic algorithms are compared with the introduced new one. The obtained results represented that the proposed algorithm could produce competitive results and find the near‐optimal solution for the BN structure learning problem.

Abstract The particle swarm optimization (PSO) and the whale optimization algorithm (WOA) are two admired optimization methods that have drawn various researchers' attention. The PSO implements some particles' intelligent movements in a search space, and the WOA is originated based on the hunting mechanism of humpback whales. The PSO and WOA have different strategies for moving towards the optimum solution. Nevertheless, both algorithms' performances encounter several problems, such as premature convergence and falling in local optimums. Several approaches have been proposed to enhance meta‐heuristic algorithms' performance, such as applying the chaotic maps, adding mathematical or stochastic operators or local searches, and hybridizing the algorithms. In this article, a new hybrid algorithm denoted as chaotic‐based hybrid whale and PSO has been presented by improving the WOA, combining it with PSO, and using the chaotic maps. The hybrid algorithm has significantly more diverse movements than both of the mentioned algorithms. Therefore, it explores different regions of a problem's search space more precisely and avoids local optima. The roulette wheel selection operator has also been applied based on their fitness value to select the proposed algorithm's search agents and exploit promising regions of the search space. In the hybrid algorithm, the chaotic maps have been applied to initialize the whales' population, particles of the particle swarm, and adjust motion parameters to increase population diversity. The multi‐swarm version of the proposed algorithm with higher performance than the single‐swarm version and other methods has been introduced in this article too. The proposed algorithms have been evaluated using 23 mathematical benchmark functions, including unimodal, multimodal, and composite functions and four engineering optimization problems. The obtained results and statistical tests prove that the proposed algorithms provide competitive solutions for most of the experiments, compared to the state‐of‐the‐art and well‐known optimization meta‐heuristic methods in terms of convergence towards the global optimum, local optima avoidance, exploration, and exploitation.

Traveling Salesman Problem (TSP) is an intricate discrete hybrid optimization problem that is categorized as an NP-Hard problem. The objective of the TSP is to find the shortest Hamilton route between cities to visit all existing cities and returning to the original city, from which the route started. Various researches have been carried out on TSP, and manifold solutions have been proposed to find the shortest route between cities, but none has been able to solve this problem completely. In this paper, a novel discrete version of the Farmland Fertility Algorithm is proposed, which uses three neighborhood searching mechanisms, and one Crossover operator. The Metropolis Acceptance Criterion has also been utilized to evade the local optimal traps. Furthermore, the Roulette Wheel selection technique is used to select neighboring mechanisms during the optimization process. Moreover, a local search mechanism is used to maximize the performance of the proposed algorithm. To prove the effectiveness of the contributions, and illustrate the efficiency of the proposed algorithm, the TSP library is evaluated on 37 data sets and compared with some well-known similar methods. The simulation results showed the superiority of the proposed algorithm against other comparative methods.

Metaheuristics play a critical role in solving optimization problems, and most of them have been inspired by the collective intelligence of natural organisms in nature. This paper proposes a new metaheuristic algorithm inspired by gorilla troops' social intelligence in nature, called Artificial Gorilla Troops Optimizer (GTO). In this algorithm, gorillas' collective life is mathematically formulated, and new mechanisms are designed to perform exploration and exploitation. To evaluate the GTO, we apply it to 52 standard benchmark functions and seven engineering problems. Friedman's test and Wilcoxon rank-sum statistical tests statistically compared the proposed method with several existing metaheuristics. The results demonstrate that the GTO performs better than comparative algorithms on most benchmark functions, particularly on high-dimensional problems. The results demonstrate that the GTO can provide superior results compared with other metaheuristics.

Abstract One of the main benefits of unsupervised learning is that there is no need for labelled data. As a method of this category, latent Dirichlet allocation (LDA) estimates the semantic relations between the words of the text effectively and can play an important role in solving various issues, including emotional analysis in combination with other parameters. In this study, three novel topic models called date sentiment LDA (DSLDA), author–date sentiment LDA (ADSLDA), and pack–author–date sentiment LDA (PADSLDA) are proposed. The proposed models extend LDA through some extra parameters such as date, author, helpfulness, sentiment, and subtopic. The proposed models use helpfulness in the Gibbs sampling algorithm. Helpfulness is a part of readers who found the review helpful. The proposed models divide the words into two categories: the words more affected by the distribution of subtopic and the words more affected by the main topic. In this study, a new concept called pack is introduced, and a new model called PADSLDA is proposed for sentiment analysis at pack level. The proposed models outperformed the baseline models because according to evaluations results, the extra parameters can appropriately affect the generating process of words in a review. Sentiment analysis at the document level, perplexity, and topic coherence are the main parameters used in the evaluations.

Nowadays, the existence of continuous optimization problems has led researchers to come up with a variety of methods to solve continues optimization problems. The metaheuristic algorithms are one of the most popular and common ways to solve continuous optimization problems. Firefly Algorithm (FA) i s a successful metaheuristic algorithm for solving continuous optimization problems; however, although this algorithm performs very well in local search, it has weaknesses and disadvantages in finding solution in global search. This problem has caused this algorithm to be trapped locally and the balance between exploration and exploitation cannot be well maintained. In this paper, three different approaches based on the Dragonfly Algorithm (DA) processes and the OBL method are proposed to improve exploration, performance, efficiency and information-sharing of the FA and to avoid the FA getting stuck in local trap. In the first proposed method (FADA), the robust processes of DA are used to improve the exploration, performance and efficiency of the FA; and the second proposed method (OFA) uses an Opposition-Based Learning (OBL) algorithm to accelerate the convergence and exploration of the FA. Finally, in the third approach, which is referred to as OFADA in this paper, a hybridization of the hybrid FADA and the OBL method is used to improve the convergence and accuracy of the FA. The three proposed methods were implemented on functions with 2, 4, 10, and 30 dimensions. The results of the implementation of these three proposed methods showed that OFADA approach outperformed the other two proposed methods and other compared metaheuristic algorithms in different dimensions. In addition, all the three proposed methods provided better results compared with other metaheuristic algorithms on small-dimensional functions. However, performance of many metaheuristic algorithms decreased with increasing the dimensions of the functions. While the three proposed methods, in particular the OFADA approach, have been able to make better converge with the higher-dimensional optimization functions toward the target in comparison with other metaheuristic algorithms, and to show a high performance.