Item Recommendation Method Based on Score Modification

Abstract

The invention provides an item recommendation method based on rating correction. The items are scored according to the user experience, the first correction is made based on linear regression, and the user-item scoring matrix is ​​obtained after the first correction. It is necessary to preserve changes in user needs and prevent overfitting, so a second correction is required to calculate the weight of the corresponding original score. Feature vectors are used to characterize active users' preferences for items. Predict unknown ratings based on the K nearest neighbors of known rated items by active users. Evaluate the results obtained accordingly. The present invention can eliminate subjective features, make ratings objective, and express user preferences more accurately. The present invention adopts secondary correction, which is more sensitive to changes in interest over time, so that the system can more accurately identify the user's current interest preference, thereby greatly improving the accuracy of the generated recommendation results.

Description

technical field [0001] The invention relates to the field of recommendation systems, in particular to an item recommendation method based on score correction. Background technique [0002] In recent years, with the information revolution brought about by online social networks and Internet technology, network information has been growing in an almost terrifying trend, and network information has swept the world like an explosion. All kinds of network information, such as news information, scientific and technological information, sports information, entertainment information, etc., began to flood us from all directions. The total amount of information generated by human beings in the past 30 years has exceeded the total amount of information in the past 5,000 years, but the price density of the information is very low, and many of them contain a lot of useless and untrue information. "Information excess" and "information chaos space", it is difficult for people to rely on t...

AI Technical Summary

Problems solved by technology

Compared with other classic recommendation systems, it has high prediction accuracy, but still has some important shortcomings that need to be improved: (1) The sparseness of data generated due to the small number of user ratings, that is, in the vast majority of data sets Data with missing or zero values

(2) The problem of scalability, that is, the problem of adapting to the continuous expansion of the system scale (3) The cold start problem of new users, the quality of the initial recommendation is poor (4) The historical data will determine the quality of the recommendation, which will reduce the timeliness of the recommendation (5) ) There are no user reviews for new projects and few projects with new user reviews cause the prediction accuracy is not high enough and other important problems need to be solved

Images