Identifying Heart Attack Risk in Vulnerable Population: A Machine Learning Approach