Machine Learning Prediction of Chronic Diabetes Based on Person's Demography and Lifestyle Information

Chronic diseases such as diabetes are prevalent globally and responsible for many deaths yearly. In addition, treatments for such chronic diseases account for a high healthcare cost. However, research has shown that diabetes can be proactively managed and prevented while lowering healthcare costs. We have mined a sample of ten million customers' 360° insight that includes behavioural, demographic, and lifestyle information, representing the state of Texas, USA, with attributes current as of late 2018. The sample, obtained from a market research data vendor, has over 1000 customer attributes consisting of behavioural, demographic, lifestyle, and, in some cases, self-reported chronic conditions such as diabetes or hypertension. In this study, we have developed a classification model to predict chronic diabetes with an accuracy of 80%. In addition, we demonstrate a use case where a large volume of customers' 360° data can be helpful to predict and hence proactively prevent and manage a person's chronic diabetes. Customer and person are both used interchangeably throughout the paper.