Neural Network and Regression Methods for Estimation of the Average Daily Temperature of Hyderabad for the Years 2018-2020

Abstract

A qualitative study on temperature distribution has been executed in Hyderabad by several researchers. This study, however, is the first attempt to study temperature distribution quantitatively. Two different methods, i.e., Artificial Neural Network (ANN) and Regression Analysis (RA), have been used to determine the average daily temperature distribution for Hyderabad, a city in Pakistan. Both the methods are used to predict the average daily temperature of the years; 2018, 2019, and 2020. In Neural Network (NN) analysis, the network was trained and validated for three years with temperature recorded from 2015-2017. With the help of training and validation parameters of the hidden layer, the average d aily temperature was predicted for 2018-2020. Based on input parameters (dew point, relative humidity, and wind speed), a multiple regression model was developed, and average daily temperature for the years 2018-2020 was predicted again. For validation of the model statistical errors, Root Mean Square Error (RMSE), Mean Absolute Error (MABE), Mean Absolute Percent Error (MAPE), and coefficient of determination are calculated. The statistical errors show that multiple regression models and neural network models provide a good prediction of temperature distribution. However, the results of the neural network are better than the regression model.