Music recommendation method and device fusing user interests, equipment and storage medium

Abstract

The invention discloses a music recommendation method and device fusing user interests, equipment and a storage medium. The method comprises the steps that a sequence length vector is calculated according to the actual length and the maximum length of a user music playing sequence; determining a portrait feature vector of the user, a short-term interest feature vector of the user and a long-term interest feature vector of the user in the music scene; determining weight vectors corresponding to the portrait feature vector of the user, the short-term interest feature vector of the user and the long-term interest feature vector of the user; based on the weight vector and the sequence length vector, performing model dynamic fusion on an initial music recall model to obtain a target music recall model; and recommending music to the user based on the target music recall model. Model dynamic fusion is performed according to the sequence length vector and the weight vector, and music library recommendation is performed based on the fused model, so that a user feels that the music library of music is rich and colorful, and final user loss is avoided.

Description

technical field [0001] The present application relates to the technical field of communication computers, and in particular, to a music recommendation method, apparatus, device and storage medium integrating user interests. Background technique [0002] For music apps, the recommendation system plays a pivotal role. An excellent recommendation system can increase the stickiness of users and bring more benefits to the app. For the music recommendation process, it is generally divided into recall, rough sorting, fine sorting and rearranging. There are four stages, among which the recall stage usually selects thousands or even tens of thousands of songs that users may like from millions or even tens of millions of candidate songs. [0003] Existing recommendation systems generally use vector retrieval in the recall phase. Specifically, long-term and short-term interests are converted into user vector ue, and each candidate song is converted into item vector ie. In the recall ph...

AI Technical Summary

Problems solved by technology

[0005] The main purpose of this application is to provide a music recommendation method, device, device and storage medium that integrates user interests, aiming to solve the technical problem that music recommendation in the prior art is likely to cause user loss

Images