Heart failure is one of the familiar diseases that needs a special care and its prediction in early stage helps to take caution of its symptoms. By the invention of machine learning, some tasks become easier to handle like expectations, classification and clusstering. The main goal of this article is to predict heart failure using the patient record. This paper focus on building more than one model to predict the prescne or absence of heart failure. These models aim to reduce the risk of death and to make the life is more adaptive with this non-cureable disease. The three models are built with different number of features. The comparisin between these models is evaluated. The first model is build with 12 features and it achived the highest accurcy 84.6% in Bagged Tree classifier, whereas the second model is exploited with 6 features, its accuracy reduced to 75.6% Cosine KNN classifier. The last model is performed around 75.9 % with just two features. This study recommend s to use the whole record fetures to obtain the highest accuracy for predicting heart failar cases.
