Social Recommendation System Using Network Embedding and Temporal Information

Recommendation systems are one of the most popular and prevalent applications of machine learning. They aim to predict user interests and recommend products or items that are most likely to be of interest to them. A social recommendation can he characterized as a recommendation with an additional input, since the online users are connected to each other over social networks and have correlations which can be used to significantly increase the efficiency of recommendations. Given a continuously evolving social network, we seek to provide accurate recommendation services to large-scale heterogeneous user networks with temporal information and perform time-based updations that capture instant interests and social evolution. For this, a model is proposed that leverages the temporal semantic effects to preserve the social relationships and user behavior sequential patterns of a heterogeneous information network. Our aim is to perform the task of recommendation in a heterogeneous user-item network setting. To capture structural and semantic correlations of a heterogeneous user-item network, we use metapath2vec which is a promising network embedding method. The results obtained from the experiments demonstrate that the system can capture both the network property and its evolution, show the role of network embedding and temporal information in improving the performance of recommendation systems to predict top-K friends and items.