Predicting human body composition using a modified adaptive genetic algorithm with a novel selection operator.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Computational Biology

Background:

• Human body composition changes are early indicators of health status, preceding disease onset.
• Accurate prediction of body composition can proactively improve individual health outcomes.
• Existing prediction models suffer from low accuracy and poor adaptability.

Purpose of the Study:

• To develop a novel, highly accurate, and efficient human body composition prediction model.
• To overcome limitations of existing models in accuracy and adaptability.
• To establish a new method for predicting body composition, using percentage of body fat (PBF) as a case study.

Main Methods:

• A new parameter selection approach combining improved RReliefF and mRMR methods was designed.
• A modified adaptive genetic algorithm (MAGA) with enhanced selection strategies was developed to calculate parameter weights.
• The MAGA model was applied to predict percentage of body fat (PBF) and compared against other algorithms.

Main Results:

• The MAGA model achieved high adaptability (0.9921), minimal mean relative error (0.05%), and a strong correlation coefficient (0.982).
• The model exhibited superior performance compared to other existing models in terms of adaptability and error reduction.
• The proposed model demonstrated efficient computation with a training time of 28.58s and a running time of 2.84s.

Conclusions:

• The MAGA-based PBF prediction model offers high accuracy and strong generalization capabilities.
• The developed model provides a novel and efficient approach for human body composition prediction.
• This method holds potential for improving health monitoring and personalized medicine.