Research On The Relationship Between Network Big Data And Movie Time Series Data Based On Machine Learning And MF-DCCA

With the rapid development of the Internet,people are willing to use media tools for entertainment and leisure,search for content of interest through search engines,and publish their personal views,experiences and opinions on social media platforms,resulting in massive amounts of online big data.These big data on the Internet largely reflect people's interests,hobbies,views and emotional tendencies,and so on.As the object of people's leisure and entertainment,the film industry is closely related to social media networks.Therefore,how to efficiently mine network big data is of great significance to the development of the movie industry.Because of the network big data is complex nonlinear time series data,MF-DCCA method and LSTM model are used,and search big data and social media big data are the two main sources of online big data.This thesis selects search engine data and Twitter happiness index to study the correlation between online big data and movie time series data.Based on the Google search engine tool,the cross-correlation between the Google search volume of the financial index search terms “DJIA”,“NASDA” and “PMI” and the North American box office was analyzed,as well as the multifractal strength between Google search volume and the box office before and after the outbreak of the COVID-19 was compared.The experimental results show that there exists multifractal cross-correlations between Google Trends and movie box office,and the multifractal is significant,and after the outbreak of the COVID-19,the multifractal increases,and financial search engine data can be effective indicators for movie box office prediction and provide references for controlling film investment risks and guiding film marketing strategies.Based on the social media platform Twitter,an empirical analysis of the non-linear relationship between Twitter daily happiness sentiment(DHS)and North American movie box office is conducted.Indicators such as Hurst index,singular index and multifractal spectrum show that the cross-correlation between DHS and movie box office is positive persistence,and the cross-correlation in the long-term is more stable than that in the short-term.Using LSTM to further verify the experimental results,the prediction results of the DHS-LSTM model are better than the benchmark model,confirming the correlation between DHS and movie box office.DHS can be used as an effective predictor of movie box office,providing a reference for selecting movie scheduling dates,and thereby improving the utilization of theater resources.In addition,the stocks of movie companies are closely related to the box office released by the company,and the relationship between the Baidu Index and the stocks of movie companies is also studied.First,based on the multifractal theory,the relationship between film company stocks and the Baidu Index of PC end,Mobile end,and “PC+Mobile” was analyzed.The models of predicting the Baidu Index and stock historical trading data were further established.The experimental results showed that the multifractal strength of the Baidu Index of Mobile end is stronger than that of PC end and the “PC+Mobile”.The Baidu Index of Mobile end has a better explanatory effect on the fluctuations in the stock market of film companies.The stock trading volume has the weakest predictability,and the Baidu Index of Mobile end has the weakest correlation with the stock trading volume.