Research On Text Sentiment Analysis Based On Ensemble Learning And Deep Learning

With the development and growth of the Internet,people are more willing to share their opinions online,such as movie reviews,e-commerce reviews,and social focus.These massive texts contain extremely high value.How to tap people's emotional behind the texts has become a research focus.Traditional methods based on sentiment lexicons can no longer be applied to massive data,and the effects of machine learning methods rely too much on artificially designed features.Therefore,this paper will study the effects of deep learning methods in sentiment analysis tasks.This paper selects two Chinese text data sets in different fields from the Internet.First,the two data sets are preprocessed and the corresponding word vector are trained.Secondly,three different deep learning models are used to verify that the use of pre-trained word vectors can improve the effect of sentiment analysis.Then,to solve the problem of single method,a Bi GRU-CNN model incorporating the attention mechanism was proposed,a comparative experiment was set up,and the reason for the poor experimental effect was analyzed.Finally,in view of the problem that the use of complex models does not improve the effect of sentiment analysis,ensemble learning is introduced into deep learning,and a bagging-based CNN model,a bagging-based LSTM model,and a bagging-based GRU model are proposed,and a comparison is set.Experiments have verified that the bagging-based deep learning model can effectively improve the effect of sentiment analysis.The bagging-based CNN model,the bagging-based LSTM model,and the bagging-based GRU model proposed in this paper have accuracy reached 95.24%,97.25%,and 97.45% on the Douban data set,and accuracy on the Amazon data set reached 94.92%,95.44% and 95.72%,which performed better than other traditional models,thus verifying the effectiveness of the model proposed in this paper.