Research Of Mobile Game Users Based On Data Mining

Mobile games have developed rapidly in the past decade.In 2019,the total global revenue of mobile games has reached US $61.7 billion,an increase of 14.8% compared with that in 2018.However,with the increasing demand of users for products,users gradually concentrate on the head products,and the market competition is more intense.Due to the current domestic game production and operation decision-making is mainly based on experience and lack of objective data analysis support,resulting in low success probability.Therefore,how to rely on scientific data support to improve product success rate is recognized as a key problem to be solved in the industry.This paper mainly studies user behavior and forms user profile through data mining technology,that is,the original data such as user operation records,consumption records,social behavior records in the game background are transformed into user attributes,user tags are generated,data sets are constructed,business models are formed,and business strategies of game operation planning are finally output.Based on the actual operation data of an information network limited liability company's H5 legendary game "chasing deer in Kyushu",this paper studies the reasons for the initial loss of users,active distribution and consumption behavior.Through the research,the reliable prediction model is generated,the operation marketing strategy is put forward and the game operation system is constructed as the methodology to optimize the conversion rate,retention rate and payment rate of the game.At present,the main problems of "chasing deer in Jiuzhou" are as follows: 1.The loss rate of registered users to active users is 70%;2.The user activity of some game modules is low;3.The uneven distribution of consumption goods leads to insufficient recovery of virtual currency.We investigated the mobile game data mining methods of Tencent TRC platform,Shenzhen intelligent and other companies,combined with the common Internet product model aarrr model and shopping basket analysis,analyzed and practiced.In the existing game operation work,we record the user's static attributes,such as user's gender,regional service,gold coin,combat power,etc.,and also record the user's dynamic behavior,such as login,logout,purchase order,novice guidance,resource loading,etc.The original data are preprocessed by aggregation sampling,dimension reduction,feature selection and feature creation.Through LR logical regression,cluster analysis,classification,association analysis and other methods,the loss prediction model,active distribution model and consumption marketing strategy of users are formed.Based on the user's game behavior dynamic data,we dig out the optimization direction of the static data of the game system and the user's potential game tendency,modify the game pertinently,give the user operation activities and preferential activities to stimulate income,which has achieved the expected effect initially,but how to form the system methodology still needs follow-up research.