Cardiovascular diseases (CVD) are the leading cause of deaths around the world. There are different risk factors for this disease. In this paper the clinical parameters are used for its detection. Nowadays Machine learning is extensively used in medicine domain. This paper presents the comparison of different classifiers used for predicting the CVD. The dataset used is famous Framingham Heart disease data set. The missing values were imputed by Proximity matrix which is the matrix of similarity. To address the issue of class imbalance, under sampling was done. Validation of the result was done with 10-fold cross validation. The different algorithms used are Logistic Regression (LR), Naïve Bayes (NB), Support Vector Machine- radial (SVM-R) and Random Forest (RF). The different evaluation parameters considered for comparison are Accuracy, Sensitivity, Precision, Specificity and AUC- ROC. Comparison shows that RF performs best with accuracy of 93.73% and sensitivity of 94.74%, followed by SVM-R with accuracy 87.45% and sensitivity as 89.09%.