Recommendation method based on standard labels and item grades

Abstract

The invention discloses a recommendation method based on standard labels and item grades. The recommendation method is characterized in that the labels are standardized, namely, user-defined labels are mapped to the standard labels clear and definite in semantics, then the standard labels are used for establishing user interest models, according to the user interest models, the similarity among users is calculated, neighboring user groups are established, and then the grades of the items to be graded by the users are predicted based on item grades of the target user and the neighboring users of the target user and an improved Slope one algorithm, so that personalized recommendation is achieved. Availability of the labels which are widely used on the Web2.0 internet and can be subjected to free defining can be obviously improved, the similarity among the users is calculated by the utilization of the user interest models based on the standard labels, the similar user groups are established for the target user, the search range of the related item grades of the target user can be shrunk, the calculation amount of the algorithm can be reduced, item grade prediction through the Slope one algorithm is improved, contributions to grade prediction are improved for the users similar in hobbies and interests, and therefore personalized recommendation quality of the internet is improved.

Description

technical field [0001] The invention relates to a solution for label standardization and automatic and rapid personalized recommendation to social network users. It is mainly used to solve the problem of how to effectively use user-defined tags and user-defined ratings for some items to make personalized recommendations, and belongs to the field of data mining technology. Background technique [0002] With the advent of the Internet age, the scale of the Internet continues to expand, and complex and diverse information floods the network, which also brings about the problem of information "overload". Users cannot quickly obtain useful resources from the excessive information, which reduces the utilization rate of information. Many useful information cannot be discovered in time or even cannot be discovered, resulting in "waste of resources". The emergence of personalized recommender system (personalized recommender system) has solved this problem well. Personalized recomme...

AI Technical Summary

Problems solved by technology

With the increase of related item rating data, calculating the difference between item j and other item ratings will be a very large overhead, which will not only affect the accuracy of the recommendation, but also have a certain impact on the calculation speed of the algorithm

[0005] (2) Users are not very targeted

[0006] To sum up, tags cannot accurately express users' interests and hobbies; although the traditional Slope one algorithm avoids calculating user similarity and reduces calculation time, it cannot avoid a huge amount of calculation and physical space consumption; the prediction accuracy of traditional Slope one There is also room for improvement

Images