Image recommendation method merging visual features and user ratings

Abstract

The invention discloses an image recommendation method merging visual features and user ratings. The method includes the first step of crawling a data set and extracting item images from the data setand a user's scoring matrix of the corresponding item images; the second step of utilizing a convolutional neural network (CNN) to extract image visual features of all the collected item images, to obtain a visual feature matrix; the third step of establishing a prediction preference model, and using an element-based alternating least squares method to update the prediction preference model; the fourth step of obtaining the user's preference values of all the item images from the final prediction preference model, sorting the preference values in descending order, selecting item images corresponding to top preference values, and recommending the item images to the user. The image recommendation method merges the visual features and the user ratings so that recommendation accuracy can be improved and personalized recommendations can be achieved.

Description

technical field [0001] The invention belongs to the technical field of image processing based on computer vision technology, and mainly relates to an image recommendation method of matrix decomposition. Background technique [0002] In recent years, with the rapid development of e-commerce, a large amount of network image data has been generated. Faced with such a large amount of image data, users hope to quickly locate the image information they are interested in, and search has become a must for this purpose. function, while search is a service request initiated by the user. In order to allow the system to actively provide services for customers, an image recommendation system has emerged, which recommends images for users by analyzing the historical data that users are interested in and the image data in the image database. The most likely image content in the database is the image content that the user is interested in, that is, the image that is closest to the image tha...

AI Technical Summary

Problems solved by technology

Moreover, this traditional image recommendation only focuses on the attributes of the item itself, and cannot take into account the user's personal preferences and interests, and cannot achieve accurate personalized recommendation.

Images