IntegrateCF: Integrating explicit and implicit feedback based on deep learning collaborative filtering algorithm

Due to the expansion of e-business, the availability of products on the internet has massively increased. Finding suitable stuff from the vast array of products available on the internet is a time-consuming task. Collaborative Filtering (CF) is the most effective recommendation method for providing users with the ability to identify relevant content and, therefore, increase engagement. However, CF has several flaws, including data sparsity and cold start problems. These are ongoing research questions that pose major hurdles to the precision of the al-gorithms. Therefore, in this work, a novel neural recommendation model is proposed based on non-independent and identically distributed (Non-IID) for CF by incorporating explicit and implicit coupling interaction. The explicit interactions consist of two models, namely Intra-coupling interactions within users and items, and Inter -coupling interactions between different users and items concerning the attributes of users and items. The Intra-coupled model learns using deep learning convolutional neural networks and is combined with the Inter-coupled model. Besides explicit coupling interactions, we present a Generalized Matrix Factorization Bias (GMFB) model that systematically trains the implicit user-item coupling. Finally, we combined with explicit and implicit coupling interactions within and between users and items accompanying the extra information about users and items under a framework called IntegrateCF. Extensive experiments on two large real-world datasets have shown that the proposed model performs better than existing methods.

Automated summary

Due to the expansion of e-business, the availability of products on the internet has increased. Collaborative Filtering (CF) is the most effective recommendation method, but it faces data sparsity and cold start problems. This study proposes a new neural recommendation model based on non-iid (Non-Independent and Identically Distributed) for CF, incorporating explicit and implicit coupling interactions, and performs better than existing methods according to experiments on large real-world datasets.