Behavior prediction method and device and related product

Abstract

The embodiment of the invention discloses a behavior prediction method and device and a related product. The invention relates to the technical field of machine learning. According to the method, a feature sequence of target content and N historical behavior sequences of a target user are obtained, a shared feature extraction result is extracted based on the N historical behavior sequences and the feature sequence of the target content, and personalized feature extraction results in one-to-one correspondence are extracted based on the N historical behavior sequences. When any behavior of the target user in the target content is predicted, a prediction result is obtained based on the previously obtained shared feature extraction result and the personalized feature extraction result of the corresponding historical behavior sequence. According to the scheme, the difference of behavior characteristics of various historical behavior sequences is considered, personalized characteristic extraction is performed in a targeted manner, and prediction of different behaviors is not executed only based on a shared characteristic extraction result any more. Even in a scene in which the relevancy of the to-be-predicted behavior is relatively low, relatively high prediction accuracy can be ensured.

Description

technical field [0001] The present application relates to the technical field of machine learning, in particular to a behavior prediction method, device and related products. Background technique [0002] With the rapid development of the Internet, the content on the Internet is becoming more and more abundant, and users can easily access various types of content on the Internet through computer equipment, such as videos, books, commodities, advertisements, etc. At present, on many Internet platforms, all kinds of content have shown a blowout trend. The operation team of the Internet platform often needs to provide users with content that they may be interested in from the massive content, so the massive content brings great challenges to the operation team. A key operation performed before content recommendation is to predict various behaviors of users on the content to be recommended (such as clicks, subscriptions, comments, favorites, etc.). This operation can be achiev...

AI Technical Summary

Problems solved by technology

[0004] combine figure 1 It is not difficult to find that in the existing technology, the input features of each task of the multi-task learning model are basically the same, which is contrary to the characteristics of the user's behavior characteristics under different tasks, resulting in low accuracy of model prediction results

This accuracy problem is more prominent in learning scenarios with low task correlation

In terms of predicting various behaviors of users, the existing multi-task learning models have been difficult to meet the accuracy requirements

Images