Comparison of Conventional Regression With Machine Learning Methods for Stroke Outcome Prediction

In performing statistical analysis in clinical trials, the conventional method is to conduct univariate analysis for variable selection, followed by multivariate logistic regression to find more details about the variables. While this approach is widely used, and does deliveruseful explanations of the data, it has several shortcomings. In this paper, we compare the results obtained with conventional methods with those of alternative machine learning algorithms in terms of their ability to predict stroke outcome of patients.We also examine the different models constructed by the algorithms, and compare them with the logistic regression model. We find that machine learning can be used to predict more precise outcomes and reveal more variables than previously known that may play an important role in determining stroke outcomes.