Value Analysis Of BiliBili Website Content Producer Based On Improved RFM Model

With the rapid development of the Internet and the update of short video technology,short video has become more and more popular,and more and more people have become the viewers,producers and publishers of short video.Among many short video platforms,Bilibili website has a large number of users and strong drainage ability.However,in terms of revenue alone,non-game business plate is obviously in a weak position in terms of revenue generating ability,and advertising revenue accounts for a very low proportion.Therefore,it is particularly important to find high-quality content producer and cooperate with them in advertising so as to increase revenue.Based on this,this paper proposes the value analysis of Bilibili website content producer based on improved RFM model,and the main research work includes:(1)Select video features and rank them by importance.First of all,according to the existing video features of Bilibili website,seven features,namely the number of playback,number of bullet screens,number of thumb up,number of coins,number of favorites,number of shares and number of comments,are subjective selected,and then the feature data and video rating data of 600 popular videos on the ranking list were climbed,and 7features were taken as input variables and video rating as output variables to establish a random forest regression model.The R~2of the regression model is large,the average relative error is small,the difference between the predicted video score and the actual video score is small,the model has a good fitting effect,and each video feature has a strong ability to explain the video score.Finally,the random forest method based on OOB was used to evaluate the feature importance,and the feature proportion was determined according to the evaluation results.(2)An improved RFM model was established according to the selected main features and their proportions.First of all,this paper takes the data of the science-automobile area of Bilibili website as an example,crawled the video information of 253 content producers,and carried out data cleaning.On the basis of the original RFM model,R was replaced as the average video interaction rate I,and M was replaced as the average thumb up rate L.The values of the three variables of the average video interaction rate I,the average video release cycle F and the average thumb up rate L of each content producers were calculated.Secondly,in order to make the measurement of different variables the same,the data is normalized,when determining the weights of I,F and L in the model,the usual hierarchical method is not used,but the entropy method is adopted to determine the weights of I,F and L,which reduces the influence of subjective factors on the weights to a certain extent.Finally,K-means clustering was used to replace the five-equal division method in the original RFM model for content producers value segmentation.According to the weighted comprehensive score of IFL,content producers value segmentation was carried out to obtain the characteristics of four types of content producers values:high value,high potential,growth and low value.Based on the final content producers classification results,from the perspective of the platform and advertisers,some measures and suggestions are put forward to provide reference for the platform to develop revenue increasing strategies.