A Systematic Review On The Prediction Of Vitamin D Deficiency And Analysis

Abstract

The majority of the human population presently suffers from a deficiency of Vitamin D. The rich diet and exposure to sunlight is the key source and the deficiency more frequently affects people during the winter season. Several disorder progressions linked with Vitamin D deficiency include Cancer, Multiple sclerosis, hypertension, rheumatoid arthritis, muscle weakness, and diabetes. The study reviews the prediction of Vitamin D deficiency on Various methods including conventional methods, statistical methods, and Machine Learning approaches by investigating the results in recent years. The main objective of the review is to analyze the methods of vitamin D prediction in various studies and the comparison of Machine Learning algorithms such as K-nearest neighbor’s(KNN), Decision Tree(DT), Random Forest(RF), Adaboost, Stochastic gradient descent(SGD), Support Vector Machine(SVM), Multilayer Perceptron(MLP), Naïve Bayes(NB), etc and to estimate the performance metrics of the algorithms includes accuracy, precision, F1-score, sensitivity, specificity, mean absolute error, mean square error were compared to identify the best-fit model and to overview the non-invasive method of prediction with Vitamin D biomarkers such as tear fluid, saliva, hair mineral and skin impedance. A Deep learning approach is also an excellent criterion to be emphasized. Concepts of wearable devices on the prediction and continuous monitoring and their contribution to non-invasive prediction are discussed. The review findings help to avoid the limitations of high cost in Vitamin D testing and aid the clinicians in providing the best treatment at the required time.