Graph Neural Network with Item Life Cycle for Social Recommendation
Jianwen Wang, Zhaogong Zhang · 2022
Recommendation system is a strategy to provide users with better decisions and help them make decisions. The sequence based recommendation system is to predict the probability of users clicking on the next item, or to score something, and help users make decisions. However, for these prediction work, most models or works in recent years only connect many models, ignoring the order of user behavior, and the location information is not taken into account.Secondly, most of the modeling is based on the interests of users, which just ignores the changes of the life cycle of the item itself. In addition, in the previous work, the model based on graph neural network obtains the potential information of users by aggregating the information of their neighbors’ nodes. However, these neighbors may not interact with users in reality and have no impact on the changes of users’ preferences. Therefore, we take the real friends of users into account and speculate the impact of friends on the changes of users’ preferences through social networks, And we also take into account the short-term and long-term interests of users, and model users at the same time. At the same time, for the item, its own life cycle is changing, and the possible changes of similar items are also similar. Aiming at this point, we capture the short-term and long-term changes of the item life cycle, and jointly predict the short-term and long-term information. A large number of experiments on real data sets show that several methods are compared to verify the effectiveness of our model.