| In the current session recommendation algorithm,it has become the mainstream to use the graph neural network method to model the session sequence.But at present,most of the session recommendation models based on graph neural network ignore the influence of the time information of the user browsing the item on the recommendation effect,and there is a problem of information dissemination of non-adjacent items.In addition,the existing models also ignore all kinds of auxiliary information in the session graph,such as item transition relations and time-order information,resulting in the inability to accurately model the session sequence.Therefore,based on the above two problems,this paper proposes two new session recommendation models,and then designs a small visual recommendation system using the model proposed above.The main work is as follows:(1)In view of the fact that the existing graph neural network session recommendation algorithm does not fully consider the time information of users browsing items,and there is the problem of information propagation of non-adjacent items,a Time information and Startransformer for Session Recommendation Algorithm(TSSR)is proposed Firstly,the complex transformation between items is effectively modeled by combining the time information of users browsing items and the graph neural network to capture the local dependency information of the session.Secondly,the Star-transformer network structure is used to alleviate the information transmission problem of non-adjacent items,and the multi-head attention mechanism is combined to effectively capture the global dependency information of the session.Finally,a gating network is applied to generate the final representation of the session by combining global dependency information and local dependency information and concatenating the reverse position information,and the user recommendation list is given.The experimental results on two public data sets show that the TSSR model is superior to the latest session recommendation model in terms of recommendation accuracy and average reci Proc.al ranking.(2)Aiming at the problem that the existing graph neural network session recommendation algorithm ignores all kinds of auxiliary information,which leads to the inability to accurately model the session sequence,a Integrate Item Transition Relations and Time-order information for Session Recommendation Algorithm(RTSR)is proposed.First,the shortest path sequence between any two nodes is obtained by using the graph network structure,which is encoded as the item transition relations between corresponding items through the gate recurrent unit(GRU),and then the global dependency information of the session is captured from the perspective of the graph by combining the self-attention mechanism.At the same time,a lossless graph coding scheme is designed to alleviate the problem of information loss in the process of session graph coding.The scheme quantifies the time-order information in the session sequence reasonably,and takes it as the weight of the edges in the session graph,and then combines the gated graph neural network to obtain the local dependency information of the session.Finally,a linear combination of global dependency information and local dependency information,combined with reverse position information,finally generates the user’s interest preferences for the item,and gives a recommendation list.The experimental results show that taking the item transition relations and time-order information of session as additional auxiliary information of session recommendation can effectively improve the recommendation performance.(3)Based on the above algorithms,a personalized e-commerce recommendation system is designed and implemented.First,the RTSR model is used as the recall layer,and the user’s session sequence is input into the recall layer to recall thousands of products,and the top-K recommended product sets are obtained according to the ranking results.Then the Deep FM model is used as the sorting layer,and the recommended product sets are input into the sorting layer to reorder them.Finally,according to the sorting results,the top-ranked product information is used as the recommendation result,and the front-end interface is used to display it to the user.According to the function display results of the system,it shows that the system can use the session sequence information of anonymous users to improve the accuracy of recommendation,so as to display products with potential value to users on the client side. |
