Weight Management Programme: Use of Machine Learning Approaches to Identify Client Outcomes

Every year, 2.8 million people die from the consequences of living with overweight or obesity. In present times, weight management programmes have become a viable option for people looking to live a healthier life, however, early dropout from these programmes are a common phenomenon, as participants often opt-out before any significant weight loss is observed. Early identification of participants of such programmes who may not obtain satisfying outcomes (in terms of engagements and weight lost) would be beneficial for the programmes providers in monitoring and providing more individual plans for their clients. This study seeks to compare, evaluate, and establish the effectiveness of various machine learning approaches for predicting dropout and weight loss from weight management programmes. The data for this research was obtained from the MoreLife case study which featured data of 8029 participants. The models were evaluated using RMSE, MAE, and R-squared metric for regression, as well as Accuracy, Precision, Recall, AUROC and Kappa score for classification. The results indicate that classification is the best approach to predicting dropout, and the Random Forest model was the best classifier with accuracy of 62% and AUROC of 65%.