Research on Personalized Recommendation Algorithms Based on Collaborative Filtering

Traditional recommendation algorithms often suffer from issues such as data sparsity, lack of memory association, and prediction accuracy. This paper proposes a collaborative filtering recommendation algorithm that integrates memory and balance factors. When analyzing user interests and preferences, there are optimal design from multiple aspects such as memory factors and novelty of objects. This paper introduces a memory weight function to reflect the temporal characteristics of user interests, and incorporates a balance factor into user similarity calculation to reduce the impact of popular items on user similarity calculation, thus improving the accuracy of user similarity calculation. Experimental validation is conducted on the MovieLens dataset using MAE as the evaluation metric. The results demonstrate significant improvements in the experimental metrics, indicating enhanced recommendation accuracy and improved resilience to data sparsity.