AI-Driven Sales Forecasting in the Gaming Industry: Machine Learning-Based Advertising Market Trend Analysis and Key Feature Mining

A sales forecasting model for the video game industry is constructed, which integrates multi-source features and trend variables in the advertisement data, and realizes time series modeling and non-linear association learning between features based on the joint structure of XGBoost and LSTM. During the construction process, feature residual distribution constraints, SHAP value decomposition mechanism and dynamic feature window strategy are introduced to improve the robustness and generalization ability of the model under high-dimensional advertising indicators. The experimental results show that the proposed model reduces the MAPE by 12.7% on average on three real datasets, and the trend variable exhibits higher stability for multi-period sales prediction.