Relevancy Model for Prediction Model Training Bias Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing prediction models are influenced by feature types in training data, which can lead to biased results, and there is a need to determine and mitigate the impact of these feature types before training to ensure accuracy and reliability.
Innovation Solution
The system uses a relevancy model to assess the influence of feature types on prediction models, updating the training data by incorporating additional datasets if a feature type has a threshold amount of influence, ensuring that the feature type's impact is reduced or eliminated, thereby improving the model's accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If training data includes feature types with high influence, then the prediction model can be trained quickly, but the model results become biased and unreliable
Solution Approach 1:
The system performs preliminary analysis of training data using a relevancy model before actual model training to identify feature types with excessive influence. This preliminary action allows the system to detect potential bias sources and adjust the training data accordingly, preventing biased results before they occur rather than detecting them after training completes
Solution Approach 2:
The relevancy model provides feedback about the influence level of different feature types in the training data. This feedback mechanism enables the system to identify which features are causing bias and allows for iterative adjustment of the training data to achieve balanced feature influence, thereby improving prediction reliability
2Reliability
If the system analyzes feature type influence using a relevancy model, then prediction reliability improves, but the complexity of the training process increases
Solution Approach 1:
The relevancy model serves as an intermediary component between the training data and the prediction model. It analyzes feature influence independently and provides guidance for data adjustment without requiring modifications to the core prediction model architecture, thereby managing complexity through a dedicated auxiliary system
3Reliability
If additional datasets are obtained to update training data, then feature type influence is reduced, but the time required for data preparation increases
Solution Approach 1:
The system performs preliminary identification of problematic feature types using the relevancy model before committing to extensive data collection. This allows for targeted updates where additional datasets are obtained only for specific features that require balancing, rather than uniformly expanding the entire training dataset, thus reducing overall data preparation time
Data Source
AI summary
In certain embodiments, training data may be generated for training a prediction model. Training data including first datasets may be obtained, where the first datasets include a plurality of feature types. A determination, via a relevancy model, based on the training data, of whether a feature type satisfies a first condition may be made. If the first condition is satisfied, one or more second datasets may be obtained to update the training data, where the second datasets include the plurality of feature types. A determination, via the relevancy model, based on the updated training data, may be made as to whether the feature type satisfies a second condition. The first and second conditions may relate to whether the feature type has a threshold amount of influence on the prediction model. If the second condition is satisfied, the updated training data may be provided to the prediction model.