Prediction Analysis Of Network Novel Value Based On Data Mining

With the explosion of the online novel adaptation in 2015,a great number of successful movies and TV dramas adapted from online novels have brought us great commercial value,Copyright operators have been looking for the purchase of copyright of online novels with commercial value,which once sparked a war of Copyrights,and even exorbitant prices,aroused widespread concern and intense discussion.In the copyright competition,in order to avoid and reduce the investment risks that must be borne by the blind bidding,an accurate forecast and estimate of the commercial value of online novels is very necessary and meaningful.Due to the fact that the current domestic research on Internet fiction focuses on qualitative research,the quantitative research and analysis of network novels are lacking,in view of this situation,this paper makes a quantitative prediction analysis of online novels from the two aspects of hidden value and heat,and provides the basis for enterprise publishers to correctly judge the commercial value of online novels.Specific research work is as follows:(1)Research on the prediction of the implied value of the network novel.First,identify variables that can reflect the implied value of network novels.According to the prediction research of film and TV series,the inspiration of the selection of the film box office and the quantity of TV play on demand is chosen respectively.In this paper,the total number of clicks of online novels is selected to reflect the implied value of network novels,and chooses the independent variable to carry on the numerical prediction research.Using the M5 model tree algorithm,linear regression algorithm,and decision table algorithm to compare predictions.The predicted results show that the M5 model tree has more accurate prediction results.At the same time,it is found that there should be a negative correlation between the recommended amount of online novels which should promote the total number of online novels.In the early period of creation,the total number of online novels has a negative correlation with the increase in the total number of online novels,and the positive correlation is only started when the number of words reaches the critical value.This paper gave a detailed description of these phenomena that do not conform to people's cognition Analysis,make a reasonable explanation.(2)Research on the heat of the network novel.Due to the limitation of using single dimension to represent the value of network novels,therefore in the context of the application of the content delivery network,this paper defines the concept of network novel's heat from three dimensions of reading basic group,reading income and reading discussion.The network novel data crawled from Qidian is preprocessed,according to the law of power law distribution data fitting,establish heat rating standards using a Bayesian network prediction model,random forest algorithm and Logistic regression.A Comparative Study of Predicting the Heat of Internet Novels.The results show that the prediction accuracy of random forest algorithm is 97.097%,the mean square error is 0.1128,the classification prediction effect is better and the error rate is lower.The prediction of the popularity of online novels not only provides decision-making basis for CDN content copy deployment and improves the quality of service of content distribution network,but also complements the limitations of characterization of network novel value from a single dimension,and more deeply and objectively identifies the value of online novels for enterprises,provide support for enterprise operators decision.