Collaborative Recommendation of E-commerce Products Based on Data Mining of Consumer Purchasing Behavior Characteristics

With the development of network economy, using e-commerce platform to establish the relationship between enterprises and consumers and enhance trust has become an important goal of enterprise management. By mining the distributed information power of consumer behavior, combining the community trust and preference similarity of user relations, we can find the nearest neighbor of the target user, predict consumer purchase behavior. Apply SMOTE-ENN method to solve the uncertainty problem to ensure the accuracy of product recommendation. The improved algorithm is compared with random forest algorithm, XG boost algorithm and light GBM algorithm. The results show that the estimation performance of random memory model is the best among the three models, and the accuracy and memory of test results exceed 0.85%. It is the second and only model of LightGBM, but there are differences in predictable performance. The practice shows that the improved recommendation algorithm has obvious advantages in accuracy, recall and F-index system.