A Predictor Generator for Healthcare Applications

Many healthcare applications now utilize machine learning algorithms to analyze a patient's history, to assist in di-agnosis, and to possibly predict the next stage of patient's health. This approach gives a computerized support for healthcare providers. However, choosing an appropriate machine learning algorithm is a daunting task for a healthcare provider, partly because of their lack of experience in using such algorithms. In this paper, we describe a tool called predictor generator which helps choosing an appropriate machine learning algorithm for a healthcare application and adjusting the parameters of the selected algorithm so that the user can get a predictor for the application. The tool provides an option to select different algorithms with different combinations of parameters and saving each of them separately so that the user can experiment each predictor and choose the appropriate one that satisfies their requirements. We have tested the application of the tool for predicting mortality of kidney patients who are on hemodialysis.