Sample Subset Optimization Techniques for Imbalanced and Ensemble Learning Problems in Bioinformatics Applications.

Area of Science:

• Machine Learning
• Bioinformatics
• Data Science

Background:

• Traditional data sampling methods often use random resampling, neglecting sample quality and usefulness.
• Machine learning tasks frequently encounter challenges like imbalanced datasets, small sample sizes, and noisy data.

Purpose of the Study:

• To introduce and apply the Sample Subset Optimization (SSO) technique for enhanced data sampling.
• To demonstrate the efficacy of SSO in addressing imbalanced and ensemble learning problems.
• To showcase the utility of SSO in complex bioinformatics applications.

Main Methods:

• Developed Sample Subset Optimization (SSO), a novel data sampling technique.
• SSO utilizes a cross-validation procedure to identify and select optimal data subsets.
• Applied SSO as an under-sampling method for imbalanced learning and as a generic ensemble technique.

Main Results:

• SSO effectively identifies highly discriminative samples in the majority class for imbalanced datasets.
• SSO facilitates the construction of robust ensemble classifiers by selecting optimized sample subsets.
• Demonstrated significant advantages of SSO on various bioinformatics tasks with prevalent data challenges.

Conclusions:

• Sample Subset Optimization (SSO) offers a superior alternative to traditional random sampling.
• SSO provides a flexible and effective approach for both imbalanced and ensemble machine learning.
• The proposed SSO techniques are highly valuable for bioinformatics research and applications.