Early Identification of Vitamin D Deficiency Risk Through Public Health Screening Data.

Area of Science:

• Endocrinology and Metabolism
• Computational Biology
• Public Health

Background:

• Metabolic syndrome increases cardiovascular disease risk.
• Vitamin D deficiency is associated with metabolic syndrome and inflammation.
• Predicting vitamin D deficiency is crucial for public health management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting vitamin D deficiency.
• To identify the most effective algorithm for vitamin D deficiency prediction in a Taiwanese population.
• To utilize data from public health check-ups for predictive modeling.

Main Methods:

• Utilized a dataset of 6,046 adults aged 30+ from Taiwanese health check-ups.
• Compared six machine learning algorithms: logistic regression, random forest, SVM, XGBoost, LightGBM, and MLP.
• Employed stratified sampling and K-fold cross-validation for model training and tuning.

Main Results:

• XGBoost achieved the highest performance in predicting vitamin D deficiency.
• The XGBoost model demonstrated excellent accuracy, F1-Score, precision, and recall.
• The study validates the utility of machine learning for identifying vitamin D deficiency.

Conclusions:

• Machine learning, particularly XGBoost, is effective for predicting vitamin D deficiency.
• This predictive capability can support early detection and management of vitamin D deficiency.
• Further research should focus on improving feature selection, handling class imbalance, and increasing dataset diversity.