The Research And Implementation Of Product Recommendation System Based On Deep Learning

With the rapid development of the Internet and mobile Internet,online shopping has become a daily shopping method for people.In today's consumer upgrade market environment,how to meet the user's requirements for product personalization and enhance the user's shopping experience is an important issue for maj or e-commerce platforms.Through data analysis of user historical shopping records,the e-commerce platform can clearly understand the user's personalized preferences for goods and periodic shopping rules,and further increase the total transaction amount of the platform.Therefore,commodity personalization recommendation and product repurchase recommendation as the main content have been researched and deep learning has been used to construct the corresponding recommendation model.The following three aspects of work have been mainly completed:(1)In order to solve the problem of information overload,traditional e-commerce platforms need to model the interests of users and personalized recommendation of goods.For the product personalization recommendation,the proposed model TransRec is based on the Transformer model and the MLP model.TransRec model is the first model for personalized recommendation by using codec structure.It realizes the mining of user's potential interest and the modeling of user interest migration through position coding and self-attention at the coding end.At the decoding end,it realizes the modeling of user interest diversity through attention.The AUC index of the TransRec model on the Books dataset and the Electronics dataset reached 0.812 and 0.756 respectively;The experimental results show that the model effect of TransRec is significantly higher than other existing models.(2)In order to improve the efficiency of consumers' shopping,it is necessary to model the users'periodic purchase behavior and recommend the products for repurchase.For product repurchase recommendation,a model based on LSTM is proposed,which can realize the automatic mining of temporal information in user's historical shopping records.And a weighted average sequence loss function is proposed for product repurchase recommendation.Combining LSTM model with TCN model,a temporal convolutional recurrent network(TCRNN)is proposed.The model has the ability to quickly retrieve global historical information and can model users'periodic purchasing behavior from multiple perspectives.Experiments show that the AUC index of TCRNN on the Instacart dataset reached 0.781;The experimental results show that the model effect of TCRNN is significantly higher than other existing models.(3)A specific product recommendation system is constructed based on Flask framework.The product recommendation system can simultaneously provide users with personalized product recommendation and product repurchase recommendation.