Machine learning framework for predicting susceptibility to obesity.

Area of Science:

• Computational biology and bioinformatics
• Machine learning applications in public health

Background:

• Obesity is a global health crisis and the fifth leading cause of death worldwide.
• Rising obesity prevalence necessitates advanced methods for early risk identification.
• Timely detection of obesity susceptibility enables proactive interventions.

Purpose of the Study:

• To introduce ObeRisk, a novel machine learning framework for predicting obesity risk.
• To develop and evaluate the entropy-controlled quantum Bat algorithm (EC-QBA) for feature selection.
• To enhance the accuracy and efficiency of obesity susceptibility prediction models.

Main Methods:

• Data preprocessing included handling null values, feature encoding, outlier removal, and normalization.
• A new feature selection method, entropy-controlled quantum Bat algorithm (EC-QBA), was developed.
• EC-QBA integrates Shannon entropy for parameter control and quantum mechanisms for local search.
• Selected features were used with various machine learning algorithms (LR, LGBM, XGB, AdaBoost, MLP, KNN, SVM) with majority voting for final prediction.

Main Results:

• The EC-QBA feature selection method demonstrated superior performance over existing techniques.
• EC-QBA achieved 96% accuracy, 96% precision, 96.5% sensitivity, and 96.25% F-measure.
• The ObeRisk framework incorporating EC-QBA significantly outperformed modern obesity prediction strategies.

Conclusions:

• The proposed ObeRisk framework with EC-QBA offers a highly accurate and effective approach for obesity risk prediction.
• EC-QBA represents a significant advancement in feature selection methodologies for complex health data.
• This framework has the potential to facilitate earlier interventions and improve public health outcomes related to obesity.