Customer Classification And Pattern Mining Algorithm In Network Game

With the Internet’s popularity in the society, online game industry places greater importance in IT company.There are some problem for game statistics mining.This article includes one model for customer classification, one new algorithm to search customers’ feature, the purpose is to do good effect for online game industry.Firstly,the article gives a RFM model for customer classification.Directly classification and K-means both can do the function.We do the experiments to test the two different methods,For less data,they have the same performance,but the direct data is reliable,more effectively to solve problem, we use the threshold RFM to module customer classification. For large data, K-means clustering algorithm has the strong ability of data acquisition,so K-means algorithm is better.Secondly,the article proposes a algorithm to search for customer feature. The algorithm is to find such model to describe the customers’feature:To find frequent pattern in the target customer database.to get unique frequent pattern measured by odds ratio.The new algorithm is better than Decision Tree and Association Rule.More patterns than C5.0,and all meaningful results.And there’s no meaningless results like Association Rule.The final part is the application for online game,including feedback mechanism which can support the game produce.Also it’s very helpful by using association purchase analysis to set the activity and items’price.In the practice,customer classification and pattern mining are integrated applied for operation progress.