Research on Game Recommendation Algorithms Based on Hybrid Models
Yonghao Wang, Chiawei Chu, Kaiye Li · 2024
Game recommendation algorithms can help users narrow down their search and find games of interest. The mainstream game recommendation algorithms currently are collaborative filtering game recommendation algorithms. However, collaborative filtering mainly relies on user ratings and similarity to make recommendations. When user data includes a substantial amount of social information, this method fails to analyze the users' social data comprehensively. It cannot grasp the users' attitudes, opinions, and emotions, leading to inaccurate recommendation results. In this paper, we propose a hybrid recommendation model. In this model, we combine collaborative filtering algorithms with sentiment analysis, taking both users' gaming data and social data into account within the hybrid model. After selecting the most appropriate K value, we use the KNN classification algorithm for personalized recommendations, optimizing the user-based collaborative filtering game recommendation algorithm. The evaluation of the recommendation results is derived from MSE, Accuracy, F1, Precision, and Recall. The results indicate that the recommendation algorithm, after hybridization, shows an improvement in recommendation accuracy compared to the user-based collaborative filtering game recommendation algorithm, proving the effectiveness of this research.