| With the continuous popularization of mobile intelligent terminals,mobile games have also developed rapidly.And the explosive growth of the mobile game market has also led to the uneven quality of mobile games in the market.Excellent products can attract a large number of users.Therefore,we propose and implemente a corresponding scheme that uses user data to improve the quality of mobile games based on data mining from the following two aspects.1.Churn preventionIn order to reduce the churn rate in mobile games,we take a casual game of mobile games that focuses on the number of active players in the game as an example.Firstly,the poten-tial relation between users'churn rate of levels and the game's own attributes is explored.Secondly,according to the relation we found,we adjust the game level properties to re-duce the users' churn rate of levels.At the same time,this paper gives a comparison of users',churn rate before and after the adjustment,which proves the effectiveness of the proposed scheme.2.Pay promotionWe also take a war-based mobile game that pays attention to in-app payment as an ex-ample to improve users' experience and promote users to purchase from multiple per-spectives.Firstly,Using the collaborative filtering algorithm based on user clustering,the user's personalized recommendation system is established based on the historical data between users and props.And through offline verification,the recall of the recommended system reaches 71.3%.so that the performance of the personalized recommendation sys-tem is verified.Secondly,using the historical data of users,we generate a prediction model of users' first payment based on ensemble learning methods.And in order to prove the performance of the model,we evaluate the model comprehensively by using evaluation index such as accuracy,F1 score and ROC curve.The data mining scheme we proposed improves the users' in-app experience as well as the quality of mobile games,and enables the relationship between mobile games and their users to be better developed. |