Forecasting Future Water Demand InBaghdad: A Comparison Between The Two Methods Multiple Regression And MLP-Neural Network Method

Abstract

The research aims to predict the quantities of water needed in the city of Baghdad for a period ofThe next 24 months were forecasted using two models: Multiple Regression and MLP-Neural Network. This model relies on the city's water consumption data obtained from the Baghdad Municipality, specifically the Baghdad Water Department, for the period from January 2014 to May 2024, a total of 125 months. It also takes into account some influencing factors such as population, maximum temperature, and monthly humidity, to estimate the quantities Baghdad will need in the future. This results in the best and most accurate model for predicting water consumption in Baghdad.

The research concluded that the best model suitable for predicting water consumption in the city of Baghdad among the models used is:MLP(3,3,1) in terms of accuracy measures and the average absolute error rate (MAPE=5%), while the average absolute error rate for the Multiple Regression model (MAPE=6%). The research also found that there is a correlation and influence between some influencing factors (population and maximum temperature) and water consumption, as these factors were able to explain 0.703% of the changes in water consumption in Baghdad. Finally, the research recommends relying on artificial intelligence models to support water planning decisions, and periodically updating water consumption data to enhance the accuracy of future predaictions. The research also recommends the necessity of taking the necessary measures to limit water consumption in the city, through pricing, awareness, education, intermittent supplies, and other methods that preserve water resources, address some influencing factors, and achieve water sustainability.