Book

Description

This work investigates the use of survival machine learning models in management for outcome prediction based on the evidence that exists in the healthcare literature. Twenty survival models, as well as over ten survival machine learning algorithms, were evaluated to determine their primary strengths and limitations.  In the first section of the research, we compare and contrast the most common models in terms of their similarities and distinctions, as well as their data types and assessment techniques. We also outline the principles that must be followed by all machine learning algorithms for survival analysis.  Using the R packages, four machine learning algorithms representing each family (trees, multi-task, kernel, and deep network) were used to evaluate the time to failure of US banks in the last 22 years, checking the purchase behavior of a set of 1000 web clients, and the bankruptcy behavior of startups in the US.  By evaluating survival, the findings demonstrate how machine learning algorithms may be used to and determine the optimal methods to use depending on more than twelve constraints, such as suppressed data, among several others.