Imbalanced Data Classification Techniques

Application of class-balancing algorithms to diverse plasma metabolomics datasets using brain tumor as an example

This study evaluates class-balancing algorithms on metabolomics datasets from brain tumor patients and controls, finding that undersampling improves model performance while oversampling risks overfitting.

Undersampling techniques for large datasets

The paper benchmarks undersampling techniques for imbalanced DNA-encoded library datasets using three machine learning models, finding that max_sim performs best and implementing the pipeline in DELight.

Binning-based Methods for Model Fitting and Data Splitting Improve Machine Learning Imbalanced Regression

This paper introduces binning-based methods for improving imbalanced regression in genomics, showing that reciprocal density weighting outperforms other approaches.

Commonly used software tools produce conflicting and overly-optimistic AUPRC values

The paper evaluates 10 software tools for computing AUPRC and finds they produce conflicting and overly-optimistic results.

The Use of Class Imbalanced Learning Methods on ULSAM Data to Predict the Case-Control Status in Genome-Wide Association Studies

This paper compares ML methods for predicting T2D status from GWAS data using imbalanced learning techniques, finding that MLP with SMOTE achieves the highest accuracy.

Class imbalance should not throw you off balance: Choosing the right classifiers and performance metrics for brain decoding with imbalanced data

This paper evaluates the impact of class imbalance on ML performance in neuroscience, showing that Accuracy is misleading and recommending Balanced Accuracy and Random Forest with stratified cross-validation.

MLcps: Machine Learning Cumulative Performance Score for classification problems

The paper introduces MLcps, a Python package that combines multiple performance metrics into a single score for classification problems, tested on biological datasets to assess model robustness.

METAbolomics data Balancing with Over-sampling Algorithms (META-BOA): an online resource for addressing class imbalance

META-BOA provides a web tool to balance metabolomics data using over-sampling algorithms and evaluates their impact through visualization and classification.

Beyond accuracy: Measures for assessing machine learning models, pitfalls and guidelines

Evaluating four families of machine learning performance metrics on simulated and public neuroimaging data shows classification accuracy performs worst for statistical power, model comparison, feature selection, and reliability.