Imbalanced-Data Model Ensembles for Lower False Positive Rates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models suffer from high false positive error rates, particularly when dealing with imbalanced data sets, which can lead to unnecessary costs and penalties.
Innovation Solution
A method involving resampling imbalanced data sets to generate multiple models with varying sampling ratios, followed by ensemble learning to combine these models using weighted averages, thereby reducing the false positive error rate.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional machine learning models are trained on imbalanced data sets, then the model can be deployed for prediction, but the false positive error rate increases
Solution Approach 1:
The patent segments the model development process into multiple stages: generating multiple data sets with different sampling ratios, training separate models on each data set, and then combining these models through ensemble learning. This segmentation allows each individual model to be simpler while the collective ensemble achieves lower false positive error rates.
Solution Approach 2:
The patent systematically varies the sampling ratio parameter across multiple data sets (e.g., different ratios of positive to negative target data points). By training models on data sets with different sampling ratios and then ensembling them, the system optimizes the false positive error rate without requiring any single model to be overly complex.
2Measurement precision
If data resampling is performed to reduce false positive error rates, then model accuracy improves, but computational time and resources increase
Solution Approach 1:
The patent performs data resampling and model training in advance during the model development phase. Multiple data sets with different sampling ratios are prepared beforehand, and models are trained on these preprocessed data sets. This preliminary action allows the system to achieve better prediction accuracy without incurring additional computational time during actual deployment.
Solution Approach 2:
The patent combines multiple models trained on different sampled data sets into an ensemble. By merging the predictions of these models through weighted averaging or voting mechanisms, the system achieves higher prediction accuracy while distributing the computational workload across multiple simpler models rather than requiring one highly complex model.
Data Source
AI summary
As described herein, a base model based on imbalanced data may be selected for a machine learning process associated with a specific application. A first false positive error rate may be generated based on the selected base model. A plurality of imbalanced data sets may be generated based on the imbalanced data associated with the base model. A plurality of models may be generated based on the generated plurality of imbalanced data sets. A subset of the outputs of the plurality of models may be ensembled and a second false positive error rate may be generated based on the ensembled output of the subset of the plurality of models. The second false positive error rate may be determined to be less than the first false positive error rate.