Research And Application Of Game Sales Forecasting Model

In the environment of fierce competition in the game market,game sales prediction is an important research direction faced by game development companies and publishers.Game sales can help game development companies and publishers better understand market demand and player behavior.Therefore,by forecasting game sales,the design and release strategies of game products can be optimized and the market competitiveness and profitability of game products can be improved.Due to the rapid development of the game industry and the explosive expansion of game data,the data sets used in relevant research are relatively complex and without much reference.At the same time,there are few researches on the prediction of game sales at home and abroad,and the prediction methods used still have some problems such as low accuracy and simple prediction algorithm.Its applicability and reliability are not guaranteed,which is not conducive to the practical application of relevant research results.Existing research on game sales prediction also focuses less on the impact of game characteristics on game sales prediction.To solve the above problems,this paper proposes a variety of game sales prediction models based on the game sales characteristic data set.The specific research contents are as follows:(1)Use the game’s characteristic data and sales data to construct the game’s sales characteristic information.Through feature selection,features with high correlation to game sales are selected to build game sales data set,so that the prediction model can learn the correlation of features and enhance the accuracy of the model in regression prediction.(2)Related prediction algorithm based on machine learning,proposed a game sales prediction model based on decision tree,GBDT and random forest.Through a variety of different machine learning prediction algorithms,the game sales prediction model is established,and the parameters of the three game sales prediction models are adjusted.According to the evaluation index,all the models have high prediction accuracy and reliability.(3)A game sales forecasting model with multi-model combination based on integrated learning Stacking is proposed.The model consists of two layers of learning model.The first layer is composed of decision tree,GBDT and random forest model,and the second layer is composed of linear regression algorithm.Through the comparison and application of model experiments,the analysis of experimental results shows that in the root mean square error(RMSE),coefficient of determination(R~2)and average absolute percentage error(MAPE),the game sales prediction model based on the combination model has better prediction effect than the single game sales prediction model.The game sales prediction model proposed in this paper has high accuracy.At the same time,the game sales prediction model based on the combination model can enhance the generalization ability of the prediction model and reduce the risk of overfitting.