Системний аналіз та прогнозування витрат води за інформацією з бази даних WISE

Abstract

This work addresses the problem of systematic analysis and forecasting of surface water flow in Europe using data from the WISE (Water Information System for Europe) monitoring database. Exploratory data analysis was performed on a dataset of over 5.7 million observations, a data preprocessing pipeline was developed, and two predictive models were built and compared — LightGBM and a deep neural network. The LightGBM model demonstrated superior accuracy (R²=0.7469, MAE=2.5464) and is recommended as the primary model for hydrological indicator forecasting.