TEMPERATURE MODELING OF RANGPUR DIVISION, BANGLADESH : A COMPARATIVE STUDY BETWEEN ARTIFICIAL NEURAL NETWORK AND LINEAR REGRESSION MODELS

The objective of this paper is to draw a comparison between the artificial neural network and linear regression models for temperature modeling in Rangpur division. Data exploited in the this paper were compiled from Bangladesh department of Meteorology. Multiple Linear Regression (MLR) model and then Artificial Neural Network (ANN) model were performed on these data sets to understand which model gives better results. For both cases MLR and ANN all of the data sets were classified as test and training case and finally checked for cross validation. The outcomes of Multiple Linear Regression model were examined in contrast to the outcome of artificial Neural Network model in case of model fitting for temperature. The ANN model functions capably relative to MLR. Independent variables like the pressure of Sea level, Humidity, Dew point temperature, Wind speed in besides Rain fall all were captured by the previously mentioned artificial Neural chain together with Multiple Linear Regression. This is probably the first attempt to model temperature using ANN and MLR for Saidpur, Dinajpur and Rangpur Stations. The greater part of studies in model fitting for climate variables are Linear or Nonlinear Regression models, but this research work takes a different approach including ANN model.