Deep Learning-based User Behavior Data Mining in Precise Recommendation of E-commerce Platforms

Abstract The continuous expansion of the e-commerce market scale and the rapid growth of the number of commodities have made how to accurately recommend commodities that meet the needs of users become the most concerned issue for e-commerce platform merchants. In this paper, through the combination of deep learning model and user’s behavioral sequence data mining, PMCA-BiLSTM is constructed as an accurate recommendation model for e-commerce platform, which is mainly composed of BiLSTM network, attention mechanism and residual convolutional neural network. On the basis of the recommendation model, this paper designs an e-commerce platform accurate recommendation system, and evaluates the performance of the system and the corresponding recommendation model. The system test results show that the longest response time for a user request is 2796ms, which is within 3 seconds, and the error rate of all test requests is 0, indicating that all the simulated requests can be correctly processed by the system, and that the system is able to give users a good user experience. The PMCA-BiLSTM model in this paper outperforms other comparative models in both HR and NDCG evaluation metrics on both Yoochoose1/64 and Diginetica datasets with different number of iterations and different Top-K, which verifies the validity of this paper’s method. The recommender system and its recommendation model designed in this paper provide a feasible path for accurately recommending goods according to user needs.