E-commerce hybrid recommendation algorithm based on user dynamic interests

Aiming at the information overload problem in e-commerce, a hybrid recommendation algorithm for e-commerce based on users' dynamic interests was studied. In the construction process of this algorithm, the user's dynamic interest factors are integrated to make the recommended results more time-effective. The combination of forgetting curve and feature construction is used to form a recommendation algorithm with dynamic characteristics. The recall rate, classification accuracy rate and F1 parameter are selected as the evaluation indicators of the model, and the indicators are compared with the LR, FR and GBDT algorithms. It can be seen from the results that, except for the classification accuracy, there is not much difference between the four algorithms, in terms of recall rate and F1 parameter, the algorithm of this study is better than the other three comparison algorithms, proving that the algorithm has good performance. It can be used to recommend products of user interest in e-commerce.