Machine Learning Prediction Modeling
The ML model was constructed using Azure Machine Learning Studio (Microsoft Corporation). The data was split into an 80/20 distribution, with 80% of data used to train the model, and the remaining 20% to test performance.12,13{Bur, 2019 #1359;Bur, 2019 #1465} Various two-class decision models were explored for development, including decision forest, decision jungle, boosted decision tree, logistic regression, and neural network. The specifications of the model were set on a parameter range in order to grant the ML the most flexibility in model development. The outcome of interest, primary treatment modality, was identified as the label.