Cardiovascular disease is one of the leading causes of death in the world. The risk of death is important to predict to determine treatment or behavior and lifestyle changes in cardiovascular patients. Medical record data of cardiovascular patients can be used as input in predicting life expectancy. This study offers the construction of a life expectancy prediction system for cardiovascular patients. Prediction using multilayer perceptron method by testing various scenarios. In addition, feature selection methods, namely correlation based filter (CBF), linear discriminant analysis (LDA), and principal component analysis (PCA) are applied to obtain relevant features to improve classification performance. Based on the experiments conducted, the average accuracy using CBF and LDA feature selection is 84% and 84.7%, respectively. In the best trial, CBF is able to produce accuracy, precision, recall, and f-measure with value of 91.7% 85% 89.5% and 87.2%. Based on these results, it can be concluded that this prediction system is able to provide fairly accurate results