A novel two-step feature selection based cost sensitive myocardial infarction prediction model

Abstract

Considering the rapid growth, complications and treatment side-effects of MI, so using data mining techniques seems necessary. On the other hand, in real-world MI cases are much less compared to healthy cases. The traditional algorithms for imbalanced problems lead to very low Sensitivity, thus, we propose a cost sensitive SMO model that utilizes a Two-Step Feature Selection, which aims to propose a model for prediction MI with regard to its imbalanced dataset to achieve a proper performance. In the dataset the MI cases in training set reduced to 9 against 410 healthy cases. After selecting 62 features, by feature selection, the Cost sensitive SMO which is allocated different misclassification cost as penalties is applied on the dataset. The results have shown positive impacts on performance.