Sentiment analysis of domestic movie reviews based on a discrete regression model

Abstract

In the context of the rapid development of big data information digitization, how to use big data technology to analyze the sentiment of domestic movie reviews is a hot topic of current concern. In this paper, we use two similarly themed Chinese movie reviews as the basis and use discrete regression models to study the emotional tendency of movie reviews. Firstly, the ROC curves of the regression model are plotted based on the convolutional neural network to construct the text classification task metrics and the logistic discrete regression algorithm. The prediction probability at the bottom of the curve is 56.3%, and the 95% confidence interval is [56%, 56.6%], indicating the model’s high accuracy. Then, the factors affecting effective tendency were analyzed by the logistic discrete regression model, and both user level (UL) and per capita disposable income (PCDI) had significant effects on affective tendency (ET) (both p-values were less than 0.001), and the coefficients of both influencing factors were less than 0. This indicates that user level (UL) and per capita disposable income (PCDI) have significant negative effects on affective tendency (ET) have a significant negative effect on affective tendency (ET), i.e., the higher the user level, the more likely it is to have a bad review, i.e., the higher the per capita disposable income, the more likely it is to have a bad review. This study provides a comprehensive and objective evaluation of domestic movie reviews and provides a guiding reference basis for the development direction and progress of Chinese movies.

Keywords

• big data
• convolutional neural network
• measurement index
• discrete regression model
• domestic movie review
• 62g08