Recommendation model training method and device

Abstract

The invention discloses a recommendation model training method and device, and relates to the technical field of computers. A specific embodiment of the method comprises the steps of performing offline calculation on bottom layer data to obtain a target attribute value; according to the target attribute value and a preset recall strategy, calculating a recommendation result in real time based on an attention mechanism and displaying the recommendation result; calculating a recommendation result conversion rate according to the behavior of the user on the recommendation result; and according to the recommendation result conversion rate, performing parameter adjustment based on a group optimization strategy to train a recommendation model. According to the embodiment, the recommendation model can be trained through real-time online learning, the rule that the user preferences change is learned in time, the interpretability of the recommendation model is higher, the recommendation result coverage is comprehensive, the difference of recommendation strategies of different types of shops can be reflected, cross-shop data can be recommended, and the recommendation result is more scientific.

Description

Technical field [0001] The present invention relates to computer technology, and in particular relates to a method and apparatus for recommending training model. Background technique [0002] Existing recommendation systems are mostly based on historical data of past user behavior, training model recommended by the predictive model is offline, and then use the recommended model for the user to personalize recommendations. But for each store e-commerce platform, the personalized recommendation is different from the home page of the shop's recommendation, its small amount of user access, data sparse, and are independent shops, the number of users is much larger than the number of goods in the shops, so most common recommendation algorithm does not apply personalized recommendation shop. [0003] In implementing the invention, the inventors have found that there is at least the following problems in the prior art: [0004] (1) recommended offline model predicts metastasis can not me...

AI Technical Summary

Problems solved by technology

[0004] (1) The recommendation model uses offline prediction, which cannot adapt to the transfer of user preferences;

[0005] (2) The calculation of scoring preference only considers the role of a single behavior, making the coverage of recommendation results not comprehensive enough;

[0006] (3) When different types of stores use the same recommendation model, it cannot reflect the differences in the recommendation strategies of different types of stores;

[0007] (4) The recommendation model of each store only recommends within the store, and cannot recommend cross-store data;

[0008] (5) The recommendation model uses a single algorithm, and the recommendation results have certain defects and limitations

Images