Sentiment Tendency Analysis Of Online Movie Comments Based On Machine Learning

When a new film has being released,lots of movie reviews can be seen on maoyan,douban and other film communication platforms immediately.Obviously,these comments can influence the box office of a movie.‘Hello,Mrs.Money' is such a movie that the attitudes of the users in maoyan and douban turned out to be opposite.So,what's the actual view of the audiences for these movies?The web clawler technology and the Sentiment tendency analysis can solve such problems.Now Web crawler technology can help us get and collect film reviews on douban and maoyan.Sentiment tendency analysis is one of the hottest research fields of artificial intelligence,which help us analyze the sentiment tendency of user's comments and predict the user's attitude by Machine Learning.The paper mainly implemented the sentiment analysis module for the film reviews of Hello,Mrs.Money on douban and maoyan.Firstly,the paper analysed those Machine Learning algorithms which are commonly used in sentiment tendency analysis,such as naive bayes algorithm and support vector machine,and introduced their basic principles,studying the advantages and disadvantages of these algorithms.Secondly,a web crawler module was designed to get tens of thousands of film comments from two movie-communication platforms whose names are maoyan and douban.Furthermore,this thesis also conceived a computer program to washed and filtered out those comments which are of low quality,such as content repetition or less information.A dictionary was made for improving the precision of word segmentation,greatly reducing the appearance of ambiguity words.Chi-square statistics and TF-IDF method are used to select the features of the film reviews.This paper also showed the performance of support vector machine and naive bayes algorithm in this data set,which can help us found the optimal solution,then we found the best classifier for this data set :naive bayes classifier.In order to improve the precision of svm classifier,this thesis also explained how to get the best hyper parameters in detail.In addition,this paper also used simple random sampling to obtain a sample to represent the whole data set,using naive bayes classifier to estimate the praise rate of the film,and analyzed the difference between the male audiences and the female audiences in sentiment tendency.This paper concluded that ‘Hello,Mrs.Money ' is not such a movie that deserved to pay for.what's more,female audiences liked the film more than male audiences.