Collaborative Filtering Recommendation Algorithm Based on K-Means and GCN

In the internet age, various contents flood people’s internet life, causing information redundancy, so performing more useful information extraction becomes an important task. Among the recommendation algorithms, the most common one is the collaborative filtering algorithm, which has the problem of data sparsity when performing matrix construction due to the poor relationship between users and items, which affects the effectiveness of recommendations. To address the data sparsity problem, the thesis proposes a collaborative filtering recommendation algorithm (KGCF) based on K-Means and GCN, which introduces K-Means and GCN, using the ability of K-Means to aggregate data and the ability of GCN to extract features in non-Euclidean space to obtain the hidden relationships between users and items, and populate the similarity matrix of users and items to alleviate the The paper uses the MovieLens dataset to improve the recommendation performance of traditional collaborative filtering algorithms. The paper uses the MovieLens dataset for comparison experiments, and uses MAE as the evaluation metric. The results show that this paper’s algorithm is better than similar algorithms in solving the sparsity of collaborative filtering data.