Weighted Majority Voting for Imbalanced Binary Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning methods struggle to effectively handle imbalanced datasets, particularly in applications like fraud detection, medical diagnosis, and predictive maintenance, where the number of positive examples is significantly smaller than negative examples, leading to high false positive rates.
Innovation Solution
An ensemble learning method that involves generatively oversampling minority class examples using kNN and multivariate probability distributions, learning base classifiers on subsamples, and combining them with a diversity-aware weighted majority vote classifier to minimize diversity on positive samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard classification methods are used on imbalanced datasets, then the training process is simple, but the false positive rate is high and classification precision deteriorates
Solution Approach 1:
The patent segments the imbalanced dataset into multiple balanced subsamples through stratified sampling, where each subsample contains equal representation of minority and majority classes. This segmentation allows standard classification algorithms to be trained on balanced data, improving precision while maintaining reliability without requiring complex algorithms.
Solution Approach 2:
The patent performs preliminary data preprocessing by generating synthetic minority class samples using SMOTE (Synthetic Minority Over-sampling Technique) before training the classification model. This preliminary action balances the dataset distribution, enabling standard classifiers to achieve high precision and low false positive rates without algorithmic modifications.
2Quantity of substance
If synthetic minority class examples are generated through oversampling, then the class balance is improved, but the risk of overfitting and loss of data diversity increases
Solution Approach 1:
The patent uses SMOTE to generate synthetic copies of minority class examples by interpolating between existing minority samples and their nearest neighbors. This copying approach creates realistic synthetic data that maintains the underlying data distribution and diversity, improving class balance while reducing overfitting risk compared to simple duplication methods.
Solution Approach 2:
The patent applies parameter changes by using k-nearest neighbors (with configurable k values) and interpolation parameters in the SMOTE process. These parameter adjustments control the degree of synthetic sample generation and diversity, allowing optimization between class balance improvement and overfitting prevention.
3Measurement precision
If ensemble methods are used to improve classification performance, then the accuracy is improved, but the computational complexity and training time increase
Solution Approach 1:
The patent segments the training process into independent parallel training jobs, where multiple classification models are trained separately on different balanced subsamples. This segmentation enables efficient parallel computation and reduces overall training time while maintaining the benefits of ensemble methods for improved accuracy.
Solution Approach 2:
The patent combines multiple independently trained classification models into an ensemble system that aggregates their predictions. This merging approach leverages the diversity of models trained on different subsamples to improve overall classification accuracy while managing complexity through modular architecture.
4Ease of manufacture
If the dataset is heavily imbalanced with few positive examples, then the data collection is easier, but the detection of positive cases becomes unreliable
Solution Approach 1:
The patent generates synthetic copies of the scarce positive examples using SMOTE, creating additional training data without requiring extensive data collection efforts. This copying approach maintains the ease of data collection while dramatically improving the reliability of positive case detection by providing sufficient training examples.
Solution Approach 2:
The patent performs preliminary synthetic data generation to balance the dataset before training detection models. This preliminary action ensures reliable positive case detection by creating sufficient training examples, avoiding the need for difficult and time-consuming data collection campaigns.
Data Source
AI summary
An ensemble learning based method is for a binary classification on an imbalanced dataset. The imbalanced dataset has a minority class comprising positive samples and a majority class comprising negative samples. The method includes: generatively oversampling the imbalanced dataset by synthetically generating minority class examples, thereby generating a generated dataset; using the generated dataset to generate subsamples, and learning a base classifier on each of the subsamples to determine a plurality of base classifiers; and learning a weighted majority vote classifier by combining outputs of the base classifiers. Each of the base classifiers is assigned a weight in such a way that a diversity between the base classifiers on the positive samples is minimized.