Recommendation algorithm for agricultural and sideline products on e-commerce platforms based on deep learning

Aiming at the problem of low accuracy in e-commerce platform recommendation of agricultural and sideline products, an agricultural and sideline product recommendation algorithm based on deep learning was studied. Fully analyze the problems existing in the process of recommending agricultural and sideline products, study the mechanism of item-based collaborative filtering algorithm, and combine the document topic and matrix decomposition hybrid algorithm. Highlighting the advantages of both, a topic weighted fusion collaborative filtering algorithm was formed. Through experiments, this algorithm can effectively improve the recommendation effect of agricultural and sideline products on e-commerce platforms. Compared with the simple item collaborative filtering algorithm, the algorithm studied in this study can effectively improve the accuracy of recommended targets by about 5%, and also has a 2% improvement in the diversity of recommended targets. In terms of accuracy of prediction results, through experimental comparison, the root mean square error of this algorithm is also lower than other comparative algorithms, which proves the advantages of this algorithm.