Building a Cardiovascular Disease predictive model using Structural Equation Model & Fuzzy Cognitive Map

According to Public Health Agency of Canada, Cardiovascular Disease (CVD) is the leading cause of death among adult men and women. Various research works have applied machine learning/data mining algorithms to predict CVD, but these methods suffer from a) lack of transparency of the predictive model building, b) lack of capability to introduce human wisdom, and c) lack of sufficient data. In this paper we provide a novel approach to tackle these issues and design a very robust and reasonably accurate model. Our approach is based on Structural Equation Modeling (SEM) and Fuzzy Cognitive Map (FCM). We used Canadian Community Health Survey, 2012 data set to test our approach. The designed model has 79% area under the ROC curve and 74% accuracy. We have used only the 20 most significant attributes, but we believe that adding more attributes and having an expert heart specialist panel would further improve the accuracy of the system.