R-Ensembler: A greedy rough set based ensemble attribute selection algorithm with kNN imputation for classification...

Area of Science:

• Machine Learning
• Data Mining
• Bioinformatics

Background:

• High-dimensional medical datasets present challenges like the curse of dimensionality, missing values, and irrelevant attributes.
• Preprocessing is crucial for machine learning models to avoid computational inefficiency and performance degradation.

Purpose of the Study:

• To propose R-Ensembler, a parameter-free greedy ensemble attribute selection method.
• To enhance the prediction of diseases by selecting optimal attribute subsets from medical data.

Main Methods:

• Utilizes rough set theory with attribute-class, attribute-significance, and attribute-attribute relevance measures.
• Employs k-nearest neighbor (kNN) imputation for missing value treatment.
• Combines multiple attribute subsets using a novel n-set intersection method to reduce bias.

Main Results:

• R-Ensembler demonstrated superior performance compared to five state-of-the-art attribute selection algorithms on seven benchmark medical datasets.
• Statistical significance of improved attribute subsets was confirmed using paired t-tests with Naïve Bayes, decision trees, and random forest classifiers.

Conclusions:

• The R-Ensembler method is highly effective in selecting relevant, significant, and non-redundant attributes.
• This leads to improved performance in machine learning tasks for medical datasets.