A Novel Reformed Reduced Kernel Extreme Learning Machine with RELIEF-F for Classification

Area of Science:

• Machine Learning
• Data Science
• Biomedical Engineering

Background:

• Large-scale data processing presents challenges for traditional algorithms.
• Reduced Kernel Extreme Learning Machine (Reduced-KELM) shows promise but suffers from prediction instability and data redundancy.
• Existing methods struggle with the computational complexity of large datasets.

Purpose of the Study:

• To propose a novel model, Reformed Reduced Kernel Extreme Learning Machine with RELIEF-F (R-RKELM), for improved human activity recognition.
• To address the limitations of Reduced-KELM, including prediction instability and computational complexity.
• To enhance classification performance on large-scale human activity datasets.

Main Methods:

• The study proposes the Reformed Reduced Kernel Extreme Learning Machine (R-RKELM) model.
• RELIEF-F attribute selection is employed to discard irrelevant features.
• A new sample selection approach is introduced to reduce training samples and improve stability.

Main Results:

• The R-RKELM model achieved superior classification performance compared to the baseline model.
• Accuracies of 92.87% (HAPT), 92.81% (HARUS), and 86.92% (Smartphone) were obtained.
• The proposed model effectively addresses prediction instability and computational complexity.

Conclusions:

• The R-RKELM model offers a robust solution for human activity recognition with large datasets.
• The integration of RELIEF-F and a novel sample selection method enhances model efficiency and accuracy.
• This research contributes to advancements in machine learning for complex data analysis.