Recommending Costume Matching with User Preference and Expert’s Suggestion

With the continuous development of e-commerce, clothing products have been greatly expanded in terms of types, styles, colors, etc., and the quantity of clothing products is increasing. How to choose your favorite clothing from a large number of products and carry out the right and appropriate dressing has become a problem that people need to consider in their daily life. Therefore, the emergence of user-oriented recommendation system comes into being, which solves the problem of information overload. However, existing work usually focuses on forming collocation recommendations from image features and related textual descriptions, which are primarily subject to low accuracy and non-personalized issues. In this paper, we consider the clothing mix purchase forecast from the users’ point of view. Specifically, we first analyze each user’s clothing preference category from the user’s historical purchase behaviors, which is used to measure each user’s preference style. Next, we select the candicated products that match with his purchased clothing according to the experts’ suggestion. Then, we calculate the similarity score between the clothing products in users’ preference and experts’ suggestion. Finally, we sort clothing items in descending order of the similarity scores, so as to generate the final recommendation list.