Ngangaivy2501/smart-water-kiosks-demand-forecasting

Data science project for forecasting water demand and optimizing smart water kiosk refill schedules in Kilifi County Smart Water Kiosks Demand Forecasting Project Overview Optimizing Smart Water Kiosks Operations Through Demand Forecasting in Kilifi County This data science project focuses on improving the operations of smart water kiosks in Kilifi County through data analysis and demand forecasting. Smart water kiosks use prepaid cards to record water transactions, including the time of purchase and volume of water sold. However, fixed refill schedules can result in some kiosks running out of water while others still have substantial amounts remaining. This project applies data analysis and machine learning techniques to identify demand patterns and develop a model that can help predict water demand and support more efficient refill scheduling. Objectives The main objectives of this project are to: - Analyze historical smart water kiosk transaction data. - Identify patterns and trends in water demand. - Explore factors that influence water consumption. - Perform data cleaning and exploratory data analysis. - Develop and compare machine learning models for demand prediction. - Identify the most suitable model for predicting liters of water sold. - Provide insights that can support improved kiosk refill scheduling. Dataset The dataset contains information related to smart water kiosk operations and environmental conditions. Key variables include: - Tank capacity - Water level before refill - Price per liter - Total sales - Temperature - Rainfall - Truck arrival delays - Daily population served - Water quality - Hour of transaction - Kiosk location - Payment method - Refill status - Sensor status - Holiday and weekend indicators - Day of the week - Month Target Variable The final regression model uses: "liters_sold" as the target variable. Data Preparation The dataset was prepared before modeling through several preprocessing steps, including: - Checking for missing values - Identifying and handling outliers - Converting variables into appropriate data types …