A data-driven study of live e-commerce selection optimization and demand forecasting

In order to solve the problem of shortage of live broadcast e-commerce products and serious inventory backlog, the research on data driven live broadcast e-commerce product selection optimization and demand forecasting was proposed. This paper first constructs an e-commerce service balanced recommendation user ontology, calculates the semantic similarity between ontologies, obtains a set of similar users, and takes the nearest neighbor object of the predicted target user as the prediction group, calculates the trust subgroup with high prediction probability in the group, makes balanced prediction through the dynamic attitude of uncertain nearest neighbors, and completes the balanced prediction of user demand in e-commerce services. The experimental results show that, in terms of accuracy, the prediction accuracy of top-n recommendation algorithm in cloud environment, P2P network node information trust calculation model simulation research, short-term photovoltaic power generation prediction algorithm based on fuzzy hyperbolic tangent model is about 69.17%, 71.13%, 76.31, while the method in this paper is about 95.32%. In terms of prediction reliability, the top-n recommendation algorithm P2P network node information trust calculation model simulation research, short-term photovoltaic power generation prediction algorithm based on the fuzzy hyperbolic tangent model, the prediction accuracy is about 62.91%, 65.28%, 73.12%, while this method is about 94.38%. Conclusion: This method can effectively improve the accuracy of user demand prediction in e-commerce services.