Exponential piece wise regression for rainfall data

Abstract

There are several regression models for the rainfall in the literature. Some of them are linear regression model, multiple regression model, piece wise regression model, Poisson regression model, negative binomial regression model, etc. Various piece wise regression models are linear piece wise regression, polynomial piece wise regression, cubic piece wise regression, quadratic piece wise regression, etc. This paper proposes three piece wise regression models using exponential smoothing models, i.e, simple exponential smoothing spline, holt’s exponential spline, and damped exponential spline for annual rainfall data from 1970 to 2017 with two knots at 1990 and 2005 with three subdivisions. Symmetric mean average percentage error is used as an accuracy measure for choosing the best model among the proposed three exponential spline models.