Content Recommendation Model Sampling for Active User Bias
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Content recommendation models are biased towards active users, neglecting the interests of low activity users and degrading relevance metrics for other content recommendations.
Innovation Solution
Implement sampling algorithms to modify streaming event data sets by limiting or capping the sampling numbers and rates of very active users and content categories, and dynamically modifying sampling weights to reduce bias in content recommendation models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If streaming events from active users are used as training data for content recommendation models, then active users can find items of interest more easily and stream more, but the model becomes biased towards active users and neglects low activity users' interests
Solution Approach 1:
The patent applies parameter changes by modifying the sampling rates of different user groups in the training data. Specifically, it reduces the sampling rate for active users and increases the sampling rate for low activity users, thereby changing the statistical parameters of the training data distribution to achieve a more balanced model that serves all user segments equally well
2Reliability
If sampling algorithms are implemented to reduce bias from very active users, then bias in content recommendation models is reduced and underrepresented groups are given equal consideration, but the complexity of the training process increases
Solution Approach 1:
The patent implements parameter changes through configurable sampling rates and weights that can be adjusted based on user activity levels. By changing these parameters dynamically during data processing, the system achieves bias reduction without requiring complex algorithmic structures, thus balancing model reliability with implementation simplicity
Data Source
AI summary
Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for reducing active user or active content category bias in content recommendation systems. An example embodiment operates by modifying a streaming event data set by selecting a voting algorithm. The voting algorithm reduces an impact of highly occurring data points by sampling the streaming event data set to generate a sampled streaming event data set, wherein the highly occurring data points comprise data points generated by the active users or the active content categories. The embodiment further trains, by a machine learning engine and based on the sampled streaming event data set, a machine learning model to generate a reduced bias content recommendation model and generates, based on the reduced bias content recommendation model, content recommendations for subsequent selection and rendering on a media device.