A hybrid ensemble and evolutionary algorithm for imbalanced classification and its application on bioinformatics.

Area of Science:

• Data Mining
• Machine Learning
• Bioinformatics

Background:

• Imbalanced data classification is a core challenge in data mining.
• Existing methods like random under-sampling and ensemble learning have limitations.
• Random under-sampling may discard crucial data, while ensemble methods often ignore data correlations.

Purpose of the Study:

• To introduce a novel Hybrid Adaptive sampling with Bagging Classifier (HABC) for improved imbalanced data classification.
• To address the shortcomings of existing sampling and ensemble techniques.

Main Methods:

• Calculated adaptive sampling rates based on dataset characteristics.
• Applied density-based under-sampling and over-sampling.
• Integrated sampled data subsets with a Bagging classifier for prediction.
• Utilized multi-objective particle swarm optimization to enhance prediction accuracy.

Main Results:

• The proposed HABC method demonstrated superior performance compared to state-of-the-art algorithms.
• Experiments were conducted on benchmark datasets (UCI, KEEL) and bioinformatics datasets.
• HABC effectively handles imbalanced data by preserving representative samples and utilizing data correlations.

Conclusions:

• HABC offers a robust and effective solution for imbalanced data classification problems.
• The adaptive sampling strategy combined with ensemble learning and optimization significantly improves classification outcomes.
• The method shows promise for applications in data mining and bioinformatics.