JMIR Publications

ABSTRACT

Background:

Overweight or obesity (OW/OB) is a primary health concern that significantly impact non-communicable disease burdens and threatens the national productivity and economic growth. Given the large complexity of obesity etiology, machine learning (ML) algorithms offer a promising alternative approach in disentangling interdependent factors for OW/OB prediction.

Objective:

This study examined the performance of three ML algorithms and compared them with logistic regression (LR) to predict OW/OB among working adults in Malaysia.

Methods:

Using the data of 16,860 participants (mean age 34.2 ± 9.0 years, 41% males, 41.8% OW/OB) from Malaysia’s Healthiest Workplace survey by AIA Vitality 2019, predictor variables comprising sociodemographic, job characteristics, health and weight perception and lifestyle-related factors were modeled using the Extreme Grading Boosting (XGBoost), Random Forest (RF), Support Vector Machine (SVM) algorithms and LR to predict OW/OB based on a body mass index cut-off of 25.

Results:

The Area under the Receiver Operating Characteristic curve (AUC) were 0.81 (95% confidence interval, CI 0.80, 0.82), 0.80 (95% CI 0.79, 0.81), 0.80 (95% CI 0.78, 0.81) and 0.78 (95% CI 0.77, 0.80) for XGBoost, RF, SVM, and LR models, respectively. Weight satisfaction was the top predictor, together with ethnicity, age and gender as the consistent OW/OB predictor variables for all models.

Conclusions:

Based on multi-domain online workplace survey data, this study produced predictive models that identified OW/OB with moderate-to-high accuracy. The performance of both ML-based and logistic regression models were comparable when predicting obesity among working adults in Malaysia.