A Product Similarity Method Based on Deep Confidence Network

Abstract

To improve product recommendation network, this paper mainly proposes a product similarity calculation algorithm based on deep confidence network. A high-dimensional product is firstly constructed and then input into the DBN model to obtain low-dimensional product feature data. Founded on the low-dimensional product feature data, the similarity between products can be calculated by the cosine formula. Through the data experiment, it is found that as the output dimension of the low-dimensional product feature matrix decreases, the similarity of the product similarity matrix also decreases, which means that the information extracted from the original input matrix is refined, and the effective information of a product is increasing, which means that the information extracted from the original input matrix is refined, and the effective information of a product is increasing.