There are two common challenges in particle swarm optimization (PSO) research, that is, selecting proper exemplars and designing an efficient learning model for a particle. In this article, we propose a triple archives PSO (TAPSO), in which particles in three archives are used to deal with the above two challenges. First, particles who have better fitness (i.e., elites) are recorded in one archive while other particles who offer faster progress, called profiteers in this article, are saved in another archive. Second, when breeding each dimension of a potential exemplar for a particle, we choose a pair of elite and profiteer from corresponding archives as two parents to generate the dimension value by ordinary genetic operators. Third, each particle carries out a specific learning model according to the fitness of its potential exemplars. Furthermore, there is no acceleration coefficient in TAPSO aiming to simplify the learning models. Finally, if an exemplar has excellent performance, it will be regarded as an outstanding exemplar and saved in the third archive, which can be reused by inferior particles aiming to enhance the exploitation and to save computing resources. The experimental results and comparisons between TAPSO and other eight PSOs on 30 benchmark functions and four real applications suggest that TAPSO attains very promising performance in different types of functions, contributing to both higher solution accuracy and faster convergence speed. Furthermore, the effectiveness and efficiency of these new proposed strategies are discussed based on extensive experiments.

The review text received by crowdsourcing participants contains valuable knowledge, opinions, and preferences, which is an important basis for employers to make trading decisions, and crowdsourcing participants to improve service level and quality. However, there are two kinds of emotional polarity in the review text, the attention paid to sentiment classification of review text with fuzzy emotional boundaries is insufficient. This paper proposes a supervised text sentiment classification method with Latent Dirichlet Allocation (LDA) to improve the classification performance of review text with fuzzy sentiment boundaries. Taking the review text of crowdsourcing participants on the Zhubajie platform as the data set, using N-gram, Word2vec, and TF-IDF algorithms to extract text features. The LDA topic model is applied to expand the number of text features and extract eight topics that affect employers’ sentiment tendencies. Text classifiers are constructed based on Support Vector Machine (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GDBT), and Extreme Gradient Boosting (XGBoost) algorithms, and the effectiveness of the sentiment classification methods are verified by ten-fold cross-validation and confusion matrix. Experimental results show that using the LDA topic model to extend the features of review text can effectively alleviate the problem that the classifier is difficult to distinguish the sentiment categories of different emotion polarity words coexisting text, and enhance the ability of emotion boundary fuzzy text classification. Based on TF-IDF and LDA to extract and expand text features, the GBDT text sentiment classifier with the accuracy of 0.881; the F1-measure of the second, third, fourth, and fifth categories samples are 0.462, 0.571, 0.278, and 0.647 respectively, which is better than SVM, RF, and XGBoost classifiers and has the best classification performance.

Feature selection is an important pre-processing step in machine learning and data mining tasks, which improves the performance of the learning models by removing redundant and irrelevant features. Many feature selection algorithms have been widely studied, including greedy and random search approaches, to find a subset of the most important features for fulfilling a particular task (i.e., classification and regression). As a powerful swarm-based meta-heuristic method, particle swarm optimization (PSO) is reported to be suitable for optimization problems with continuous search space. However, the traditional PSO has rarely been applied to feature selection as a discrete space search problem. In this paper, a novel feature selection algorithm based on PSO with learning memory (PSO-LM) is proposed. The goal of the learning memory strategy is designed to inherit much more useful knowledge from those individuals who have higher fitness and offer faster progress, and the genetic operation is used to balance the local exploitation and the global exploration of the algorithm. Moreover, the k-nearest neighbor method is used as a classifier to evaluate the classification accuracy of a particle. The proposed method has been evaluated on some international standard data sets, and the results demonstrated its superiority compared with those wrapper-based feature selection methods.

Oversampling techniques are widely used to rebalance imbalanced datasets. However, most of the oversampling methods may introduce noise and fuzzy boundaries for dataset classification, leading to the overfitting phenomenon. To solve this problem, we propose a new method (FSDR-SMOTE) based on Random-SMOTE and Feature Standard Deviation for rebalancing imbalanced datasets. The method first removes noisy samples based on the Tukey criterion and then calculates the feature standard deviation reflecting the degree of data discretization to detect the sample location, and classifies the samples into boundary samples and safety samples. Secondly, the K-means clustering algorithm is employed to partition the minority class samples into several sub-clusters. Within each sub-cluster, new samples are generated based on random samples, boundary samples, and the corresponding sub-cluster center. The experimental results show that the average evaluation value obtained by FSDR-SMOTE is 93.31% (93.16%, and 86.53%) in terms of the F-measure (G-mean, and MCC) on the 20 benchmark datasets selected from the UCI machine learning library.