Forecasting Travel Demand When the Explanatory Variables Are Highly Correlated

Abstract
This paper discusses the problem of multicollinearity among explanatory variables commonly encountered in travel demand forecasting by using ridge regression. The authors demonstrate that when severe multicollinearity exists and the pattern of collinearity among regressors changes over time, ridge regression models yield forecasts with significantly lower forecast error than ordinary least squares models.

This publication has 0 references indexed in Scilit: