Predicting vitamin D deficiency using optimized random forest classifier

Vitamin D can be acquired from various dietary sources, but exposure to sunlight's ultraviolet rays can convert a natural compound called ergosterol present in the skin into vitamin D.

The current study aimed to investigate vital parameters and use an optimized random forest (OptRF) classifier to understand better and predict the effect of environmental and nutritional factors of Vitamin D deficiency.

A predictive, cross-sectional, and correlational design was utilized in a study involving 350 male and female Tabuk citizens in Saudi Arabia. The Weka machine-learning tool was employed for comprehensive data analysis, with the OptRF algorithm being tailored through advanced feature selection methods and meticulous hyperparameter tuning.

In addition to the OptRF classifier, a number of traditional machine learning techniques have been tested and compared on the dataset of vitamin D to analyze and build the predictive model for classifying vitamin D deficiency. In general, the OptRF-based predictive model can statistically describe data for determining significant features related to Vitamin D deficiency. OptRF demonstrated its ability to classify vitamin D deficiency cases with high accuracy 91.42 %.

This study showed that Tabuk citizens are at high risk of vitamin D deficiency especially among females (gender predictor) with little regard to age, income, smoking, and sun exposure. In addition, exercise, less Vitamin D intake, and less intake of Calcium are also predictors of Vitamin D deficiency. Due to the link between Vitamin D Deficiency and major chronic illnesses, it is important to emphasize the importance of identifying risk factors and screening for Vitamin D Deficiency. It may be appropriate for nutritionists, nurses, and physicians to promote community awareness about strategies to improve dietary Vitamin D intake or consider recommending supplements.