Risk object identification method and device, storage medium, equipment and program product
By integrating non-semantic information and behavioral pattern features, and using the LightGBM model to analyze interaction data in social software, the problem of insufficient recognition capability of semantic models when facing the rapid changes of black and gray industries is solved, achieving efficient and accurate risk identification and resource conservation.
Patent Information
- Application Number
- CN202511483581.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-16
- Publication Date
- 2026-02-10
AI Technical Summary
Existing semantic models suffer from decreased recognition capabilities when faced with rapidly evolving and modified traffic-generating texts from black and gray market activities. They are unable to adapt quickly and are easily affected by noisy data, resulting in insufficient recognition accuracy and an inability to effectively intercept high-frequency traffic-generating activities from black and gray market activities.
By integrating content pattern features based on non-semantic information with object behavior pattern features, and using machine learning models to extract and identify features from interactive data, including character type distribution, homophonic substitution, content repetition, and behavioral pattern features, a LightGBM model is constructed for risk identification.
It achieves efficient and accurate identification of risky targets in the black and gray market, improves the model's identification efficiency and accuracy, reduces dependence on labeled data and training cycles, reduces platform resource consumption, and reduces social interference to ordinary targets.
Smart Images

Figure CN121502733A_ABST
Abstract
Claims
1. A method for identifying risky objects, characterized in that, The risk object identification method includes: Retrieve interaction data sent by the target object within a preset time period; Feature extraction is performed on the interactive data to obtain content pattern features and behavioral pattern features. The content pattern features are statistical features constructed based on non-semantic information and are used to characterize the inherent information type or properties of the interactive data. The behavioral pattern features are used to characterize the dynamic activity patterns of the interactive data. The content pattern features and the behavior pattern features are input into a pre-trained machine learning model to obtain the risk identification result of the target object.
2. The risk object identification method as described in claim 1, characterized in that, The interaction data includes text data and / or non-text data. Before performing feature extraction on the interaction data to obtain content pattern features and behavior pattern features, the risk object identification method further includes: The corresponding text data obtained by converting the non-text data; Each of the text data is subjected to length normalization processing to obtain multiple content items of preset length. The normalization processing includes at least one of truncation or padding. A preset number of content items are selected in chronological order to construct an interactive data set for feature extraction.
3. The risk object identification method as described in claim 1 or 2, characterized in that, The content pattern features include at least one of the following: character type distribution statistical features, homophonic substitution statistical features, content repetition statistical features, or key segment hit statistical features. The character type distribution statistical feature is used to characterize the proportion of different character types in the interactive data; the homophonic replacement statistical feature is used to characterize the proportion of keyword replacement using homophones in the interactive data; the content repetition statistical feature is used to characterize the distribution of frequently occurring content items in the interactive data; and the key segment hit statistical feature is used to quantify the degree of matching between the interactive data and the toxic feature.
4. The risk object identification method as described in claim 3, characterized in that, The content pattern features include the character type distribution statistical features. Feature extraction is performed on the interaction data to obtain the content pattern features, including: Divide each character in the content data according to its type; The number of characters of each type is counted and the proportion of each type of character relative to the total number of characters is calculated to obtain the character type distribution statistical characteristics. The character type distribution statistical characteristics include at least one of the following: total content length, proportion of special characters, proportion of spaces, proportion of newline characters, proportion of English characters, proportion of Chinese characters, and proportion of numbers.
5. The method according to claim 3, characterized in that, The content pattern features include the homophonic substitution statistical features. Feature extraction is performed on the interactive data to obtain the content pattern features, including: Establish a mapping relationship between core keywords and their corresponding homophonic replacement characters to obtain a homophonic dictionary; Based on the homophonic dictionary, homophonic replacement characters are identified in the content data; The statistical features of homophonic substitution are obtained by calculating the total number of homophonic replacement characters or the proportion of the number of duplicate homophonic phrases identified to the total number of characters in the content.
6. The method according to claim 3, characterized in that, The content pattern features include content repetition statistics features. Feature extraction is performed on the interaction data to obtain the content pattern features, and the features also include: Count the frequency of occurrence of each content item in the content data and sort them by frequency; The top N most frequent content items are selected as target content items, and the proportion of each target content item's occurrences to the total number of content items is calculated to obtain the content repetition statistics.
7. A risk object identification device, characterized in that, The device includes: The acquisition unit is used to acquire interactive data sent by the target object within a preset time period; An extraction unit is used to extract features from the interactive data to obtain content pattern features and behavior pattern features. The content pattern features are statistical features constructed based on non-semantic information and are used to characterize the inherent information type or properties of the interactive data. The behavior pattern features are used to characterize the dynamic activity patterns of the interactive data. The identification unit is used to input the content pattern features and the behavior pattern features into a pre-trained machine learning model to obtain the risk identification result of the target object.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted for loading by a processor to perform the risk object identification method as described in any one of claims 1-6.
9. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing a computer program, and the processor executing the risk object identification method according to any one of claims 1-6 by calling the computer program stored in the memory.
10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the risk object identification method according to any one of claims 1-6.