DCRS: a deep contrast reciprocal recommender system to simultaneously capture user interest and attractiveness for online dating

Recently, the reciprocal recommendation, especially for online dating applications, has attracted increasing research attention. Different from the conventional recommendation problems, the reciprocal recommendation aims to simultaneously best match users’ mutual interests to make the recommendations. However, the most existing RRS algorithms seldom model users’ interest and attractiveness simultaneously under the high-dimensional feature space. Furthermore, the sparsity of reciprocal relations seriously deteriorates the recommendation performance. Thus, we propose a novel Deep Contrast Reciprocal Recommender System (DCRS) to address the aforementioned research issues. Particularly, we resolve the sparsity issue by introducing the reciprocal neighbors to increase the number of possible reciprocal relations. Then, a novel deep contrast neural network is then proposed to model the mutual interest by contrasting between the reciprocal and non-reciprocal relations. As a result, it was able to better identify the reciprocal relations for the latter recommendation. Extensive experiments have been evaluated on two real-world datasets, and the promising results demonstrate that the proposed DCRS is superior to both baseline and the state-of-the-art approaches.