Prediction of Type 2 Diabetes through Risk Factors using Binary Logistic Regression

Abstract

The main objective of this study is to arrive at a highly efficient prediction model for early prediction of diabetes by relying on risk factors for diabetes as predictor variables. Using a binary logistic regression model, a model was built for the data of study which taken from a sample of diabetics and non-diabetics persons. The results have shown the high ability of the binary logistic regression model in predicting the diabetes-infected persons. All indicators confirm the validity and quality of the model. The results of Chi-square test (sig=0.945>.05) indicated that the model is significant. Likewise, the results of Hosmer&Lemeshow test (sig=0.945>.05) confirm that the model represents the data very well. Classification table findings were also high, as the overall percentage for the correct classification was 91%. The significant risk factors that influential in predicting diabetes can be arranged as follows: (High blood pressure, Diabetes in the family first degree, High cholesterol, Smoking, Age above 35, Overweight and Gender). All these results confirm the quality and accuracy of this model in predicting the disease, the thing which may indicate that the model can be used as a primary tool for predicting type-2 diabetes through risk factors.