Sentiment Classification via Distributed Beetle Swarm Optimization Algorithm

Abstract

Logistic regression is commonly used in sentiment classification. Based on the limitations of traditional logistic regression classifier parameter adjustment, some parameters may not reach the global optimum, resulting in the low performance of the classifier. To solve this problem, we propose an improved beetle swarm optimization algorithm (IM-BSO) to optimize the hyperparameters of logistic regression to improve classification accuracy. The IM-BSO algorithm adopts an adaptive adjustment strategy of the inertia weight and starts a larger search at the beginning of the algorithm iteration. As the inertia weight continues to decrease, the search range of the beetle is continuously reduced. In addition, the IM-BSO algorithm uses the levy flight strategy and the Cauchy mutation strategy to increase the diversity of the beetle swarm in the later iterations and jump out of the local optimum, which improves the convergence accuracy of the algorithm. Due to the large amount of data to be processed and the long calculation time of the IM-BSO algorithm, we propose a new distributed and improved beetle swarm optimization algorithm(DIBSO), combined with logistic regression to form a new classification model: the DIBSO-LR model. Finally, use the model to classify the sentiment of the Twitter comment data set at different numbers of nodes, comparing the speedup ratio, the experimental results show that within a certain range, the larger the amount of data, the more obvious the speedup effect will be as the number of nodes increases.