Classification of high dimensional biomedical data based on feature selection using redundant removal.

Area of Science:

• Biomedical Informatics
• Machine Learning
• Data Science

Background:

• High-dimensional biomedical data present challenges due to numerous irrelevant or redundant features.
• These extraneous features negatively impact classification accuracy and machine learning efficiency.
• Accurate identification of core features is crucial for disease diagnosis assistance.

Purpose of the Study:

• To propose a novel filter feature selection algorithm, FSBRR, for high-dimensional biomedical data.
• To effectively identify and remove redundant and irrelevant features.
• To enhance classification accuracy and machine learning efficiency in biomedical data analysis.

Main Methods:

• Developed a novel filter feature selection algorithm based on redundant removal (FSBRR).
• Defined two redundant criteria: vertical relevance (feature-class) and horizontal relevance (feature-feature).
• Utilized an approximate redundancy feature framework based on mutual information (MI) for feature quantification and removal.

Main Results:

• FSBRR algorithm effectively reduces feature dimensions in high-dimensional biomedical data.
• The proposed algorithm significantly improves classification accuracy compared to typical methods.
• Controlled trials using three different classifiers validated the algorithm's effectiveness.

Conclusions:

• The FSBRR algorithm offers an effective solution for feature selection in high-dimensional biomedical data.
• FSBRR enhances both classification performance and computational efficiency.
• The algorithm demonstrates utility even with small sample datasets.