Seed Group Diffusion Using TGI and Logistic Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
When a large number of seed groups are present, the data calculation for determining target users for advertisement push becomes excessive, and the inclusion of low-quality seed users with redundancy can lead to poor advertisement effectiveness.
Innovation Solution
A seed group spreading method that utilizes a server to obtain and analyze seed and non-seed user groups, forming positive and negative sample sets, and employing a combination of Target Group Index (TGI), Information Gain (IG), and Logistic Regression (LR) models to determine suitable target features for effective seed group spreading.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a large number of seed groups are used for advertisement push, then the coverage and potential reach of advertisement push is improved, but the data calculation complexity and computational resources required increase significantly
Solution Approach 1:
The patent segments the large set of seed groups into smaller subsets based on feature similarity and user characteristics. By dividing the comprehensive seed group collection into manageable segments, the system can process each segment independently, reducing overall computational complexity while maintaining comprehensive coverage through aggregated results from multiple segments.
Solution Approach 2:
The patent performs preliminary filtering and preprocessing of seed groups before the main advertisement push operation. By pre-processing seed groups to identify and remove low-quality or redundant groups in advance, the system reduces the burden on subsequent data calculation operations, thereby lowering computational complexity while preserving the beneficial effect of using multiple seed groups.
2Quantity of substance
If low-quality seed users with redundancy are included in the seed group, then the quantity and coverage of target users is improved, but the advertisement effectiveness and quality of target user selection deteriorates
Solution Approach 1:
The patent extracts and removes low-quality and redundant seed users from the seed groups through filtering mechanisms. By taking out undesirable elements (low-quality users) while retaining the beneficial ones, the system maintains a sufficient quantity of target users while significantly improving the overall quality and effectiveness of the selected target user set for advertisement push.
Data Source
AI summary
Embodiments of this application provide a seed group spreading method performed at a server. The method includes the following steps: obtaining a positive sample set; obtaining a negative sample set; concatenating each positive sample in the positive sample set with a corresponding positive sample feature to form a positive sample feature vector, and concatenating each negative sample in the negative sample set with a corresponding negative sample feature to form a negative sample feature vector; obtaining a target sample feature set from the positive sample feature vector and the negative sample feature vector according to at least two of a target group index, an information gain, and a logistic regression model; and sending the target sample feature set to a decision end, receiving feedback information of the decision end, and determining, according to the feedback information, whether to spread the seed group.