METRO FLOW-CITY LEVEL WATER CONSUMPTION FORECASTING USING MLAND CLIMATE INDICATOR

The rapid rise of the world's urban populations and the
change in climate are creating barriers to properly managing water resources in many cities. Therefore, if we want to plan cities in a sustainable manner, distribute water efficiently, and prevent future water shortages, we must precisely forecast future water demand in our cities. This paper describes a machine-learning-based water usage
forecasting system called MetroFlow, which will forecast the amount of water consumed in a city by using a combination of historical water usage and climate factors. The model can improve its calculation in prediction by adding the environmental factors like rainfall, temperature, humidity, evaporation, and wind speed. The advanced feature engineering methods are used to analyze the short-term and
long-term patterns of water consumption. The machine learning models involved in the project are Random Forest, XGboost, Long Short Term Memory, these models are evaluated using regression metrics RMSE, MAE and MAPE. The machine learning model that provides the best predictions will be used to generate forecasts for the next 24 hours, 7 days, and 30 days. Additionally, an interactive visualization dashboard has been created using Streamlit to enable
users to visually explore water consumption trends and view water usage forecasts. This proposed project demonstrates how machine learning techniques that consider climate will enhance forecasting of urban water demand and assist with smart city resources management.