Estimating Breakpoints in Piecewise Linear Regression Using Machine Learning Methods

Nowadays there are many methods to explain non-linear behavior. However, among them, Piecewise Linear (PWL) regression has received a lot of attention in recent years. Accurate estimation of breakpoints is critical in the PWL model, which compresses nonlinear relationships at breakpoints and exhibits linear behavior between two breakpoints. The aim of this study is to determine the location of the break points in the most useful way by using decision tree regressor and k-means clustering, which are machine learning-based methods. From the experimental results, it was observed that the accuracy of the k-Means clustering method (98%) was higher than the decision tree regressor method (96%).