A Logistic Regression and Artificial Neural Network-Based Approach for Chronic Disease Prediction: A Case Study of Hypertension

Abstract

The global trend of population aging and the continuing maturity of the Internet of Things (IoT) technology drives the rapid development of health care. In the comprehensive applications of IoT technology, developing and constructing a prediction model for chronic diseases is a great improvement to healthcare technology as well as an exploration of IoT technology on the data-analysis and decision-making level. Considering that early detection, diagnosis and screening of hypertension plays a significant role in the prevention and reduction of the onset of cardiovascular diseases as well as the improvement of quality of life, it is of great value to figure out hypertension-related risk factors and further establish a model for the prediction of hypertension with the identified risk factors. Thus, in this paper, we put forward to integrate logistic regression analysis and Artificial Neural Networks (ANNs) model for the selection of risk factors and the prediction of chronic diseases by taking a case study of hypertension. First, binary logistic regression model was applied on experimental dataset collected from Behavior Risk Factor Surveillance System (BRFSS) to select factors statistically significant to hypertension in terms of the pre-defined p-value. Then, a Multi-Layer Perception (MLP) neural network model with Back Propagation (BP) algorithm was constructed and trained for the prediction of hypertension with the selected risk factors as inputs to ANNs. Experimental results showed that our proposed approach achieved more than 72% prediction accuracy acceptable in the diagnosis of hypertension and that the Area Under the receiver-operator Curve (AUC) was more than 0.77. The results indicate that integration of logistic regression and artificial neural networks provides us an effective method in the selection of risk factors and the prediction of hypertension, as well as a general approach for the prediction of other chronic diseases.