Forecasting Daily Discharge in Kievlom Dam Using Data Mining Techniques

Abstract

This research aims to 1) study the appropriate data mining techniques to forecast the amount of water in the dam, and 2) compare the actual water volume and the prediction with the amount of water in KiewLom Dam, Lampang Province, using data mining techniques. This research manages existing water data that is a factor in water level changes consist of the amount of water flowing into the dam, the amount of water in the dam, emissions and evaporated water.  The data were collected daily from the year 2535 – 2559 totally 25 years with 9,125 records to use for forecasting with Techniques.

            The results showed that: 1) appropriate data mining techniques used to forecast the water level in the dam consist of 4 techniques were (Regression Analysis: RA) , (Artificial Neural Network : ANN), (M5P Model Tree : M5P), and (Support Vector Machine : SVM)  by choosing from the most relevant research works, and 2) comparison results of the prediction of monthly water Level  in KiewLom dam, Lampang, using data mining four techniques found that M5P Model Tree had Mean absolute Error as low as 10.58 and was the most appropriate way to develop a system for forecasting the water in the dam. Considering the absolute value of a Mean absolute Error from each technique in ascending order was: how the model tree M5p is equal 10.58, support vector machine has an estimated value of 19.90, regression analysis method is equal 20.00, and artificial neural networks is equivalent to 21.65.