A hybrid feature selection algorithm using simplified swarm optimization for body fat prediction.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Data Science

Background:

• Obesity poses a significant global health threat, necessitating accurate body fat percentage measurement for effective treatment.
• Current methods for measuring body fat are often expensive, complex, or inaccurate.
• There is a need for practical, cost-effective solutions to accurately assess body fat.

Purpose of the Study:

• To develop and evaluate a novel hybrid feature selection method for accurate body fat percentage prediction.
• To address the limitations of existing body fat measurement techniques.
• To improve the accuracy and reduce the error in body fat prediction models.

Main Methods:

• A two-phase approach was employed, starting with a VIKOR-based multi-filter ensemble technique (VMFET) for feature filtering.
• The second phase involved an improved simplified swarm optimization (iSSO) with specific enhancements for prediction improvement.
• The iSSO incorporated a biased random initialization, an effect-based feature pruning scheme, and multiple linear regression.

Main Results:

• Extensive experiments were conducted on nine datasets to validate the proposed method's performance.
• The proposed hybrid feature selection model demonstrated a promising ability to predict body fat percentage.
• Empirical results confirmed the effectiveness of the method when compared to existing approaches.

Conclusions:

• The developed hybrid feature selection model is an effective tool for predicting body fat percentage.
• The method significantly enhances prediction accuracy and lowers prediction error.
• This approach offers a practical and cost-effective solution for body fat assessment.