Prediction And Analysis Of Influencing Factors Of Online Drama Attention

Under the background of Internet era,the popularity of mobile devices such as smart phones and tablet computers makes drama works move from traditional TV stations to network video platforms.With the advantages of low access threshold,short production cycle and high return on investment,the online drama has rapidly occupied the key market of TV industry,and its market potential is still very large.For online dramas,to catch the audience's attention is to catch the market demand,and the focus is naturally the level of attention of online dramas.A popular online play is sought after by the public,and has won the reputation of the industry,which can eventually bring rich economic benefits for the online play side.In the past,the broadcast volume was usually used as an indicator to measure the audience flow level.However,i QIYI,Youku and other platforms have canceled the display function of the broadcast volume.As a result,we can't use the video broadcast volume to reflect the attention of online drama.It has become an urgent problem to predict the performance of an online play scientifically.Through the collection,processing and analysis of online drama information,this paper studies the current situation of online drama market and the influencing factors of online drama attention,establishes the prediction model of online drama attention,and puts forward corresponding suggestions for online drama parties.First of all,we use Python to obtain information about online dramas from Baidu Encyclopedia,Sina Weibo and Douban platforms.After data cleaning,505 completed online drama works from 2014 to 2020 are selected as the research samples.For online drama data,using factor analysis to extract comprehensive indicators to measure online drama attention.Secondly,from the perspective of basic attribute information,network marketing information and short comment text information,this paper constructs features,and mines the factors that affect the attention of online drama through descriptive analysis.In view of the fact that the short review text with reference value to the audience can not directly reflect its influence,this paper creatively extracts the emotional score features to join the research by constructing the emotional analysis method of dictionary.Thirdly,we use the analysis of variance and other methods to preliminarily screen the characteristics and construct the influence index system of online drama attention.Finally,the selected 20 features are included in the model as independent variables,and three online drama attention prediction models,Lasso,support vector machine and random forest,are fitted respectively.The analysis shows that the number of performers,the total number of microblogs and the number of forwarding are the most important factors affecting the online drama attention.In order to compare the prediction effect of the model,the models before and after adding emotional features are compared and analyzed.The results show that the root mean square error of the model after adding emotional features is relatively small,which indicates that the emotional feature value extracted from the short review text is effective in predicting the attention of online drama.It's consistent with the statement that people tend to pay more attention to the works with good reputation.Secondly,the five fold cross validation method is used to compare the model with emotional features.The results show that the random forest is more generalization than the others.