Machine learning algorithms for prediction of double burden of malnutrition among Tunisian adults

Abstract

Abstract Background Malnutrition, referring to both nutritional deficits and excess adiposity (EA), currently affects one third of the world population. In Tunisia, nutritional deficits are still major challenges in a context of epidemiological and nutritional transition. The objectives of our study were to identify the potential risk factors of DBM and propose a better machine learning (ML) based model for predicting DBM among Tunisian adults. Methods About 7963 respondents aged 20 years and over were taken from Tunisian Health Examination Survey (THES), a cross-sectional national household survey undertaken in 2016. DBM was defined as the association of EA and anemia. EA was defined as a waist-to-height ratio ≥0.6 and /or overweight: BMI ≥25kg/m2. Anemia was defined as hemoglobin (Hb)<13g/L in men and Hb < 12g/L in women. The potential risk factors for malnutrition were extracted using logistic regression (LR). Five ML algorithms including Naïve Bayes, support vector machine (SVM), artificial neural network (ANN), AdaBoost and random forest (RF) were employed for predicting DBM and their performance was evaluated using accuracy, precision, recall, and area under the curve (AUC). Results LR illustrated that age, sex, region, wealth index, educational level and marital status were identified as potential risk factors for DBM. AdaBoost had the highest accuracy (89.8%), followed by SVM (89.6%). Naïve Bayes had the highest Recall (98.1%) and the most performant AUC (91.4%), while RF had the highest Precision (70.8%). The ANN had an accuracy of 89%, a precision of 70.4%, a recall of 78.7% and an AUC of 0.851. Conclusions The study demonstrated the efficacy of ML algorithms in predicting DBM in Tunisian adults. AdaBoost and SVM showed the highest accuracy, and could be utilized as an early detection tool to manage DBM and alleviate the burden on the healthcare system. However, larger and more diverse datasets are needed for further validation. Key messages • Double burden of malnutrition, defined as the coexistence of undernutrition and overnutrition, is a significant issue globally and a major challenge in Tunisia. • Machine learning algorithms, particularly AdaBoost and SVM, can be effective tools for predicting and managing double burden of malnutrition in Tunisian adults.