Hybrid Method of Logistic Regression and Data Envelopment Analysis for Event Prediction: A Case Study (Stroke Disease)

Abstract Predictive analytics is an area of statistics that deals with extracting information from data and using it to predict trends and behavior patterns. Many mathematical modeling has been developed and used for prediction, and in some cases, they have been found to be very strong and reliable. This paper studies different mathematical and statistical approaches for events prediction. The main goal of this research is to design and construct a hybrid prediction method for events prediction, based on Logistic Regression (LR) method and Data Envelopment Analysis (DEA) technique. In this study, a novel hybrid algorithm was developed, and considering the kind of collected data, the LR method was applied for input selection, and the capability of the additive (ADD) model of DEA was examined to predict the occurrence or non-occurrence of the events. To apply the proposed approach, stroke disease was selected for the case study. The results showed that any patient who was placed on the frontier has had a stroke by one or more risk factors. On the other hand, the observations that were not on the frontier had not suffered from a stroke. The overall accuracy of 88.5 percentages was obtained for the developed method.