Non-linear Variable Structure Regression (VSR) and its application in time-series forecasting

Variable Structure Regression (VSR) is a new kind of non-linear regression model, which simultaneously determines the exact mathematical structure of non-linear regressors and how many regressors there are, thereby freeing the end user from trial and error time-consuming studies to determine these. The results are based on an iterative procedure for optimizing parameters and automatically identifying the structure of the VSR model. A novel feature of this new model is it not only uses a linguistic term for a variable but it also uses the complement of that term. It also provides the end user with a physical understanding of the regressors. A Monte Carlo study shows the practical accuracy of VSR model on the classical Gas Furnace time-series prediction problem. VSR ranked #1 compared to five other methods.