Random RotBoost: An Ensemble Classification Method Based on Rotation Forest and AdaBoost in Random Subsets and Its...

Area of Science:

• Machine Learning
• Medical Informatics
• Data Science

Background:

• The medical industry generates vast datasets with numerous features daily.
• Increasing feature numbers in medical data leads to higher computational costs during inference.
• Existing methods like Principal Component Analysis (PCA) in tree-based models offer partial solutions.

Purpose of the Study:

• To propose an enhanced ensemble classification method, Random RotBoost, to address computational challenges in high-dimensional medical data.
• To improve the robustness and reduce overfitting in medical data classification tasks.
• To enhance the quality of clinical decisions through improved data analysis.

Main Methods:

• Developed Random RotBoost, an ensemble classification method utilizing an AdaBoost mechanism.
• Implemented random, automatically generated rotation subsets, replacing manual feature subset selection.
• Employed multiple AdaBoost-based classifiers within the ensemble framework.

Main Results:

• Random RotBoost demonstrated superior classification performance compared to existing methods on real-world medical datasets.
• The automated random rotation process efficiently managed high-dimensional data.
• The ensemble approach effectively mitigated overfitting, enhancing model robustness.

Conclusions:

• Random RotBoost offers a computationally efficient and robust solution for high-dimensional medical data classification.
• The method has the potential to significantly support and enhance clinical decision-making.
• This approach advances the application of machine learning in medical informatics.