Research On Commodity Recommendation Algorithm Based On Session

With the rapid development of the Internet industry,in today’s society,people’s purchase of goods has shifted from the traditional offline purchase to online purchase,which is naturally inseparable from commodity websites.Facing the exponential growth of commodities,how to efficiently screen commodities for the users of commodity websites has become the primary problem to be solved.How to accurately recommend goods on commodity websites so that users can obtain Xinyi goods has become the goal of the research.The conversational recommendation algorithm studied in this paper is actually to predict the items users click next by analyzing the user’s session click data,so as to recommend the items users are interested in to users.Because the user’s behavior is difficult to predict,in order to better analyze the user’s behavior,this paper designs the session recommendation algorithm around the cyclic neural network,and uses the memory enhancement network to increase the ability of the algorithm for long-term preference recommendation and dynamic learning.Specifically: 1)the mainstream session algorithm only depicts the user’s short-term preference and makes recommendation based on it.Lacking the description of users’ long-term preferences,the session recommendation algorithm proposed in this paper captures users’ long-term preferences based on memory enhancement network,and fuses users’ long-term and short-term preferences based on gating mechanism,so as to form long-term recommendations.2)current session algorithms are evaluated in static environment,which is rarely seen in real life,The session recommendation algorithm should learn the newly generated items and preferences and dynamically recommend users.The session recommendation algorithm proposed in this paper is based on the readability and writing characteristics of memory enhancement network,which can learn the newly generated items and preferences in the session,so as to achieve the effect of dynamic learning,At the same time,the inverted table and product quantization are used to store the data of memory enhancement network efficiently.So this paper proposes a session based recommendation algorithm.Finally,this paper uses two shopping data sets,YOOCHOSE and DIGINETICA and uses HR and MRR as evaluation indexes to verify the algorithm.The experimental results show that the results of this algorithm for conversational data sets are significantly improved compared with other conversational recommendation algorithms.It can not only provide users with accurate recommended goods,but also show a good ability to learn new sessions,which verifies the feasibility and effectiveness of the algorithm.