Research On The On-Demand Forecast Model Of Variety Shows

In recent years,in addition to the rapid development of the film and TV industry,the variety industry is also making rapid progress.For a variety show,the amount of on-demand is an important indicator to measure the broadcast effect and commercial value of the show,and it is also an important indicator for investors to decide whether to invest by considering various factors and considering risks and benefits.To a great extent,predicting the on-demand quantity of variety shows can reduce the investment risk for investors,and adjust the strategy in the process of program production,marketing and distribution in order to make correct decisions and maximize the income of investors.In this study,154 variety shows broadcast in China from 2018 to 2022 were taken as the research object,and based on the development characteristics of domestic variety shows,a forecast system of variety shows’ on-demand quantity was established with the indicators of the number of broadcast platforms,whether there is a sequel,the highest popularity,news articles,the first broadcast year,broadcast time,holiday file,duration,program duration,Douban rating,the number of people scoring Douban and the program type.Firstly,a stepwise weighted regression model is established by using R language.Six variables,namely,maximum popularity,holiday file,number of broadcast platforms,duration days,broadcast time and sequel,are gradually selected into the model,and the regression coefficients all pass the significance test,so we infer that these six factors have a significant impact on the on-demand quantity.Secondly,the random forest RF model is established by using MATLAB,and the sample data are trained and the important variables that affect the program on demand are obtained.The order of importance of each attribute is: the highest popularity,the number of douban raters,the number of broadcast platforms,the cultivation of talent show,broadcast time,news articles,cultural interviews,duration,first broadcast year,music performance category,whether it is a sequel or not,and other types.Finally,the BP neural network prediction model is established by using MATLAB again.While establishing the prediction model,this paper randomly selects ten variety shows to predict their on-demand quantity,and uses MSE and MAPE to evaluate the prediction accuracy of the model.Empirical analysis shows that the minimum MSE and MAPE values of BP neural network model are only 2.888 and 10.088,followed by random forest RF model with MSE and MAPE values of 9.507 and 12.848 respectively,and stepwise weighted regression model with MSE and MAPE values of 40.579 and 26.671 respectively.Comprehensive comparison shows that the prediction effect of BP neural network model is better than that of random forest RF model,and the prediction effect of random forest RF model is better than that of stepwise weighted regression model.Finally,according to the analysis results,some reasonable suggestions are put forward for the producers,publishers and investors of variety shows,such as making programs with sequels,moderately marketing to improve the popularity of programs,and focusing on a certain platform as much as possible.