| With the rise of social networks,viewers no longer choose movies based solely on their preferences,they are more influenced by movie reviews.This paper explores the influence of film reviews on the box office of films.Based on the traditional box office prediction model,it improves the box office prediction model and introduces the emotional characteristics of film reviews.The paper selects film reviews of domestic films from 2020 to 2022 for research,and constructs an sentiment analysis model based on the BiLSTM model through data cleaning,word segmentation,and removing stop words to fully explore the audience’s true perception of the film.The trained model achieved an accuracy of 85.6%and a comprehensive evaluation index of 0.87.indicating good classification performance.In the study of box office prediction model,this paper selects the influencing factors such as film duration,actor,director,film type,film format and distribution company,and introduces the emotional characteristics of film reviews to build a random forest,XGBoost and BP neural network for comparative analysis.The experimental results show that under different feature selection,the accuracy of BP neural network is the highest,and its performance is better than that of random forest and XGBoost models.Among them,the performance of BP neural network model based on random forest feature selection is better than that based on XGBoost model feature selection.Based on the indicators selected in the paper,it is effective to introduce emotional characteristics into the prediction model.Emotional characteristics have different influences on the three models,which have the greatest influence on random forest,followed by the influence on BP neural network.However,after removing emotional features,the RMSE values of the different BP neural networks constructed in the paper increased by more than 22.0%on the test set,and emotional features have a significant impact on the predictive performance of the BP neural network. |
PDF Full Text Request