Research on Game Interaction Modeling Based on Artificial Intelligence and Feature Extraction

This paper proposes a time series analysis algorithm based on spatiotemporal attention mechanism, aiming to significantly improve the ability to accurately model game interaction behaviors. By constructing a spatiotemporal joint attention mechanism, the algorithm can dynamically capture the complex temporal associations and spatial characteristics between players and the environment in the game scene. Compared with traditional time series modeling methods, the algorithm in this paper particularly emphasizes the in-depth mining of dynamic behavior patterns, not only focusing on the temporal changes of player behavior, but also making full use of spatial information to enhance modeling accuracy. The real scene of an MMO is chosen as a test platform, and the performance of this algorithm is evaluated by simulation. The experiment results indicate that the proposed algorithm improves significantly on several key performance indexes, and the precision of prediction is 95.3%, and the average time series modeling error is 3.7<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>. The proposed method has excellent applicability in highly complex game interaction environments. In addition, this paper further explores the influence of different key parameters in the feature extraction process through sensitivity analysis, and verifies the robustness and universality of the algorithm. Appropriate parameter adjustment can not only optimize model performance, but also maintain high stability in a wider range of scenarios.