INTEREST PREFERENCE MINING METHOD BASED ON VIEWING BEHAVIOR OF ONLINE USER

摘要

An interest preference mining method based on viewing behavior of an online user, comprising the steps of: S10, obtaining score data of a user for a viewing product according to viewing product selection behavior of the user; S20, filtering out a low-score data set according to the score data of the user for the viewing product; S30, using 90% of a high-score data set as a training set, and using 10% of the high-score data set as a test set; S40, training the training set, and estimating personalized parameters of each user; S50, predicting, according to the personalized parameters of each user, preference values of the users for unselected viewing products; and S60, sorting the unselected viewing products in descending order according to the preference values, and selecting a plurality of viewing products located at the top of the list as recommendations for the user. With the present method, the interest preference requirements for the online user viewing behavior may be met to the maximum extent.