The work uses HTCondor distributed platform, DAGMan as the Workflow Management Engine, and NFS as the distirubted file system, to parallellize the machine learning model training and evaluation work.
The machine learning models involved in this work are K-Nearest Neighbour (KNN), Random Forest, and Support Vector Machine (SVM), written in Python.
The models and application were deployed on AWS EC2.