Abnormal song identification method and device, computer device and storage medium
By analyzing song feature data using pre-trained decision tree and isolation tree models, and identifying abnormal songs by utilizing search path length, the problem of low identification efficiency in existing technologies is solved, and efficient screening of abnormal songs is achieved.
CN118113903BActive Publication Date: 2026-06-12TENCENT MUSIC ENTERTAINMENT TECH (SHENZHEN) CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TENCENT MUSIC ENTERTAINMENT TECH (SHENZHEN) CO LTD
- Filing Date
- 2024-01-25
- Publication Date
- 2026-06-12
AI Technical Summary
Technical Problem
Existing technologies are inefficient and prone to omissions when identifying black market songs on music platforms, making it difficult to effectively screen for abnormal songs.
Method used
A pre-trained decision tree model is used to classify song feature data using an isolated tree model. The search path length and standardized path length of the song feature data in the decision tree are used to determine whether a song is an anomalous song.
Benefits of technology
It improves the efficiency of identifying abnormal songs, accurately identifies and filters abnormal songs on music platforms, and reduces omissions.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN118113903B_ABST
Abstract
The application relates to an abnormal song identification method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining a pre-trained decision tree model; the pre-trained decision tree model comprises at least one isolated tree model; each isolated tree model corresponds to a song feature dimension, and the isolated tree model is a binary search tree obtained by classifying a sample song set according to song feature data in the corresponding song feature dimension; according to song feature data of a to-be-identified song, the corresponding search path of the to-be-identified song in each isolated tree model is determined; the song feature data pointed to by the search path in the isolated tree model is matched with the song feature data of the to-be-identified song in the same song feature dimension; and according to the path length of each search path, the song identification result of whether the to-be-identified song is an abnormal song is determined. The method can accurately identify and screen abnormal songs in a song database, and improves the identification efficiency of abnormal songs.
Need to check novelty before this filing date? Find Prior Art