In this work, a novel method for selecting the optimal set of input features for classifying the presence of congestive heart failure (CHF) using a supervised machine learning approach is presented. A random forest classifier (RFC) was utilized to carry out the binary classification task and two different models were explored. We employed the embedded RFC feature importance attribute for the first model, and a multi-classifier technique which integrates the feature weight importance correlation (F-WIC) method was adopted for the second model. Our results show that the second model using the F-WIC method offers superior performance (100% accuracy) and provides a generalized approach to feature engineering for machine learning models irrespective of the algorithm used. This work offers a novel method for selecting the optimal set of input features for classifying the presence of congestive heart failure (CHF) using a machine learning approach.