A framework for predicting employee health risks using ensemble model

Abstract

Through the phenomenon of data, big data and data analytics have provided an opportunity to collect, store, process, analyze and visualize an immense amount of information. Healthcare is recognized as one of the most information-intensive sectors. An urge to explore analytics has been sparked by the rapid growth of data within the healthcare sector. Most employers in Malaysia provide medical benefits that are included in the medical insurance plan for their employees. Data collected such as the history of medical claims are stored with the HR (Human Resource) which contributes to the potential of analyzing and recognizing trends within medical claims to better understand the use and overall health of the employee population. Patients with higher risk will generally convert into patients with high costs. Hence, early intervention of these patients will allow employers to potentially minimize costs and plan preventative steps. In predictive analysis, Decision Trees and Regression are typical techniques applied. The proposed framework combines an ensemble technique known as Stacking. As opposed to a single predictive model, an ensemble predictive model would yield better performance and accuracy. The objective of this paper is therefore to review current practices and past research within the healthcare sector while suggesting a practical framework for classification ensemble modeling. Preliminary findings indicated that an ensemble model can produce higher predictive accuracy and performance than a single model.