Research On User Behavior Segmentation And Interest Modeling Based On Temporal Graph Model

In the modern society,recommendation system has been widely used.From small communities to large e-commerce sites,the recommendation system is undoubtedly playing a very important role.Improving the accuracy of recommendation algorithm is always the most challenging and meaningful task.In the field of recommendation systems,how to calculate the degree of interest of an item is a critical issue.However,the user's interest is time-varying,which brings great challenges to the recommended system.Therefore,in the aspect of user modeling,how to take into account the time factor is particularly important.The session-based temporal graph introduces time information by adding a new node type named session nodes in the graph.Based on session-based temporal graph,this paper proposes a time segmentation method based on a greedy strategy,and proposes an interest decay model based on the item-session binomial graph.The main research work of this paper is as follows:1)The representation of time information in graph model is studied,and a time segmentation method based on greedy strategy is proposed,which considers different behavior intervals of different users.2)The user interest modeling and recommendation in the graph model are studied.The interest decay model based on the item-session binomial graph is proposed,which considers interest evolution over time.3)On the basis of the previous two steps,the interest decay model on the item-session binomial graph is implemented,and the recommendation method based on user collaborative filter and session-based temporal graph is implemented.And test and compare the data sets on the dataset MovieLens and Douban.The results verify the effectiveness of the work done in this paper.