Research On Deep Collaborative Filtering Recommendation Algorithm Fusing Explicit And Implicit Features

Personalised recommendation systems play a pivotal role in alleviating information overload,with collaborative filtering being one of the most widely used and successful approaches.However,existing collaborative filtering recommendation methods based on deep learning generally have certain limitations: On the one hand,most of them mostly fail to take note of the coupling relationship and non-independent homogeneous distribution nature between users and items;on the other hand,these methods fail to adequately consider the complementary relationship between explicit and implicit associations.To address the above issues,this paper proposes a deep collaborative filtering recommendation algorithm fusing explicit and implicit features.Firstly,a double-channel convolutional neural network model is constructed to extract the low-order and high-order features by using the 1DCNN channel and the hybrid CNN channel,and a mixture of 1DCNN and 2DCNN is used to learn the explicit feature interactions between users and items.Secondly,a deep latent feature representation network is constructed to learn the implicit feature interactions between users and items by using the non-linear structure of the multi-layer perceptron.Finally a combination of explicit and implicit user/item vectors are jointly trained to derive the probability of user/item interaction.The method uses a combination of explicit and implicit feedback to predict user preferences,and embeds user and item features into collaborative filtering to learn the coupling between them.A full experimental comparison between the proposed method and several other advanced methods was conducted on the Movie Lens 1M and Tafeng public datasets,and the model performance was evaluated by using two evaluation metrics,HR and NDCG.The Experimental results show that the method proposed in this paper achieved the best recommendation results on both datasets,where HR@10 reached 92.27% and 76.12%,and NDCG@10 reached 76.84% and 51.45%.The method proposed in this paper effectively improves the hit rate and accuracy of recommendations,and can alleviate the data sparsity and cold start problems in recommendation systems to a certain extent.The paper has 20 figures,12 tables,and 54 references.