Chinese Sentiment Analysis Using Regularized Extreme Learning Machine and Stochastic Optimization

In Chinese sentiment analysis, accuracy and execution speed are a comprehensive performance evaluation of an algorithm. This paper proposes an efficient method for sentiment analysis. Firstly, the classification data set is directly used to construct the corpus and word2vec for conversion of processed corpus. Secondly, regularized extreme learning machine(RLM) for sentiment analysis. Considering that the single objective only considers one objective (precision), the particle swarm algorithm is used to solve a multi-objective function for the non-dominated Pareto front optimal value. Experiments were performed on the ChnSentiCorp, and compared with other mainstream models, better classification results and faster execution speed were achieved.