RAMOBoost: Ranked Minority Oversampling in Boosting.

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Learning from imbalanced data is crucial due to widespread applications.
• Complex characteristics of imbalanced data challenge existing learning solutions.
• Robust efficiency in learning-based applications remains a significant hurdle.

Purpose of the Study:

• To introduce Ranked Minority Oversampling in Boosting (RAMOBoost) for improved imbalanced data learning.
• To address the limitations of current methods in handling imbalanced datasets.
• To enhance the robustness and efficiency of ensemble learning systems.

Main Methods:

• RAMOBoost employs adaptive synthetic data generation within an ensemble learning framework.
• It adaptively ranks minority class instances using a sampling probability distribution.
• A hypothesis assessment procedure shifts decision boundaries for difficult-to-learn instances.

Main Results:

• Simulation analysis on 19 real-world datasets demonstrates RAMOBoost's effectiveness.
• Performance was evaluated using metrics like accuracy, precision, recall, F-measure, G-mean, and ROC analysis.
• The method shows significant improvements in learning from imbalanced data.

Conclusions:

• RAMOBoost offers an effective approach to tackle challenges in imbalanced data learning.
• The adaptive ranking and synthetic data generation contribute to robust model performance.
• This technique provides a valuable solution for real-world applications with imbalanced datasets.