Sentiment Analysis Of Movie Reviews Based On FV-SA-SVM

In recent years,with the rapid development of China's economy,China has already become the second largest economy in the world.People's living standards are constantly improving,and people's ways of enjoying life are becoming more and more diversified.Watching movies is one of the main forms.As audiences have soared,so has the market.According to the State Film Administration,China's box office totaled 60.976 billion yuan in 2018,making it the world's second-largest movie market.In the 21 st century,Internet technology is changing with each passing day.With the popularity of mobile phones and numerous movie-watching apps,people can buy movie tickets and post comments online anytime and anywhere.Through the user's comment text data,we can know their emotional tendency,further analyze the audience's views on movies,and summarize the advantages and disadvantages,so as to guide the film and TV industry to develop in a better direction and maximize the value of the user's comment text data.This article crawls the film reviews of eight films in four categories: act ion,comedy,youth and suspense from the Cat's Eye APP.First,the film reviews are pre-processed,and then FV-SA-SVM is used to divide the film reviews into positive and negative reviews The results show that the accuracy of the FV-SA-SVM algorithm reaches 97.8%,95.3%,96.1%,and 97.4%,respectively.Then compare this classification algorithm with SA-SVM algorithm and traditional classification algorithm,and find that the accuracy,precision,recall and F1-Score of FV-SA-SVM algorithm are better than SA-SVM algorithm and Traditional classification algorithm,thus verifying the superior performance of FV-SA-SVM algorithm in the sentiment classification of movie reviews.Then applied this model to the emotion classification of the three action movies of "Shanghai Fortress","War Wolf 2" and "Red Sea Action".The classification accuracy rates were 94.7%,93.4% and 92.9%,which were better than SA-SVM and traditional classification algorithms.Next,the sentiment dictionary score method is used to obtain the sentiment value of the movie review for statistical analysis.Then,the semantic network analysis of the movie review is carried out to construct a semantic network to realize data visualization.Then the subject of the movie review is mined to extract the most frequent occurrences.The first four themes,mining movie features;Finally,k-Means algorithm is used for cluster analysis.By comparing the clustering results,it is found that the audiences most concerned about the movie special effects and acting actions are the two aspects of the action movie.The low score action movie is because These two aspects are not good,and finally this article puts forward relevant suggestions based on these two aspects.The conclusion of this paper is that it can help film producers understand the audience's needs and improve their own film works,and also help other viewers choose whether to watch this movie.