Multiple Relevant Feature Ensemble Selection Based on Multilayer Co-Evolutionary Consensus MapReduce.

Area of Science:

• Data Mining
• Machine Learning
• Pattern Recognition
• Bioinformatics

Background:

• Feature selection for large datasets is challenging due to noise, uncertainty, and the need to integrate multiple feature sources.
• Existing methods often struggle with the complexity and scale of modern datasets, limiting their effectiveness in real-world applications.

Purpose of the Study:

• To propose a novel Multiple Relevant Feature Ensemble Selection (MRFES) algorithm for large-scale datasets with multiple relevant feature sources.
• To develop an effective Multilayer Co-evolutionary Consensus MapReduce (MCCM) model for feature ensemble selection.
• To enhance accuracy, efficiency, and interpretability in feature selection for complex data.

Main Methods:

• Developed a Multilayer Co-evolutionary Consensus MapReduce (MCCM) model for feature ensemble selection.
• Implemented a Multiple Relevant Feature Ensemble Selection (MRFES) algorithm based on the MCCM model.
• Explored unified consistency aggregation and mechanisms for detecting noncooperative co-evolutionary behaviors to achieve Nash equilibrium.

Main Results:

• MRFES demonstrated effectiveness in solving large-scale dataset problems with complex noise and multiple relevant feature sources on benchmark datasets.
• The algorithm significantly improved the selection of relevant feature subsets, enhancing accuracy, efficiency, and interpretability.
• MRFES successfully applied to human cerebral cortex-based classification prediction, showing scalability for complex brain data.

Conclusions:

• MRFES provides an effective solution for feature ensemble selection in large-scale, complex datasets.
• The MCCM model facilitates cooperative feature selection and decision agreement among co-evolutionary processes.
• The algorithm's successful application in brain data classification highlights its potential for complex biological data analysis.