基于局部敏感哈希聚类的双通道自编码器-协同过滤混合推荐算法OA

To address the data-sparsity and cold-start problems in recommender systems,this paper proposed a dual-channel AutoEncoder-collaborative filtering hybrid algorithm based on locality-sensitive hashing(LSH)clustering.Firstly,it exploited LSH to cluster users efficiently,cutting computational overhead.Next,it designed a dual-channel AutoEncoder to jointly learn deep representations of two heterogeneous information sources—user rating behavior and movie genre preferences.Finally,a hybrid collaborative-filtering mechanism refined the predicted scores.Experiments on MovieLens-100K,MovieLens-1 M,and Amazon-Book show that the proposed model achieves NDCG@5 values of 0.371 4,0.568 9,and 0.737,respectively—up to 16％higher than the best existing methods.RMSE for rating prediction drops to 0.326 6,0.443 5,and 0.560 6,an im-provement of over 69.6％compared with the AutoRec baseline.The model outperforms competitors in recommendation accu-racy,ranking quality,and coverage,demonstrating its scalability and robustness in large-scale,highly sparse scenarios.