Dynamic Outlier Bias Reduction in Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models often suffer from bias due to outliers in training data, which can propagate and affect prediction accuracy, and existing methods lack efficient mechanisms to dynamically address and reduce this bias.
Innovation Solution
The implementation of dynamic outlier bias reduction methods, where a processor iteratively refines model parameters by identifying and removing outliers based on error sets and bias criteria, generating a non-outlier data set, and training an outlier classifier to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are trained on complete training data sets including outliers, then the model can learn from all available data, but prediction accuracy deteriorates due to bias propagation from outliers
Solution Approach 1:
The patent applies preliminary action by performing outlier detection and removal before the final model training phase. The system iteratively trains models, identifies outliers based on prediction errors, removes them from the training set, and retrains. This preliminary cleaning action prevents outlier bias from propagating into the final model, thereby improving prediction accuracy while maintaining reliability.
2Reliability
If iterative outlier removal processes are implemented to reduce bias, then prediction accuracy improves, but computational time and processing complexity increase
Solution Approach 1:
The patent implements dynamics by making the outlier detection and removal process adaptive and iterative rather than static. The system dynamically adjusts the training process by repeatedly: (1) training a model on current data, (2) identifying outliers based on error thresholds, (3) removing outliers, and (4) retraining. This dynamic iteration continues until convergence criteria are met, balancing accuracy improvement with computational efficiency by stopping when further iterations yield diminishing returns.
3Measurement precision
If dynamic outlier detection and removal methods are applied, then model bias is reduced and prediction accuracy enhances, but system complexity and processing requirements increase
Solution Approach 1:
The patent applies segmentation by dividing the machine learning pipeline into distinct modular components: (1) initial model training module, (2) outlier detection module that compares predictions against thresholds, (3) data filtering module that removes identified outliers, and (4) iterative retraining module. This segmentation allows each component to be independently optimized and managed, reducing overall system complexity while maintaining the benefits of dynamic outlier removal for improved measurement precision.
Data Source
AI summary
Systems and methods include processors for receiving training data for a user activity; receiving bias criteria; determining a set of model parameters for a machine learning model including: (1) applying the machine learning model to the training data; (2) generating model prediction errors; (3) generating a data selection vector to identify non-outlier target variables based on the model prediction errors; (4) utilizing the data selection vector to generate a non-outlier data set; (5) determining updated model parameters based on the non-outlier data set; and (6) repeating steps (1)-(5) until a censoring performance termination criterion is satisfied; training classifier model parameters for an outlier classifier machine learning model; applying the outlier classifier machine learning model to activity-related data to determine non-outlier activity-related data; and applying the machine learning model to the non-outlier activity-related data to predict future activity-related attributes for the user activity.