APPLICATION OF MACHINE LEARNING ALGORITHMS FOR DAILY WATER LEVEL PREDICTION IN KAZAKHSTAN RESERVOIRS

Abstract

It's very important to guess how high water will get for control of water, mostly in Kazakhstan. Reservoirs there are very important for watering crops, power, and giving water to people. Traditional forecasting methods often struggle to capture the complex interactions between natural processes and human activities, especially when data is limited or conditions change rapidly. This study uses three types of baseline machine learning models - Decision Tree, Random Forest, and K-Nearest Neighbors and hybrid model based on Random Forest Regressor and Multilayer Perceptron - to figure out daily water amounts using old info from Kazhydromet. The info was cleaned up, filling in missing parts, and making it all the same scale for studying. After testing different methods, hybrid model worked best. It was able to guess amounts right and did so all the time without big changes. What this work shows is that machine learning, mostly hybrid model, can give a needed assist to handling of water in Kazakhstan. This assist will provide people in charge with better ability to deal with environmental worries and social needs. The ability to guess ahead of time lets water managers make clever choices. These approaches account for data limitations and complex factors that influence water behavior. This leads to using water in a way that is more strong, fair and gets ready for the future.