Hubungan Panjang Seri Data Historik Terhadap Kualitas Data Hasil Prediksi Pada Penerapan Model Rantai Markov Untuk Peramalan Aliran Sungai

Abstract

Markov Chain Model is a stochastic model for forecasting the river flow which in his analysis always involves a long series of historical data. In most studies the method is still highly theoretical and not fully applicable significantly due to the limited data in the field.This study is an attempt to optimize the application of Markov Chain Model for its functionality extensively to extrapolate data streams. The scope of this research is basically conducted a study on the relationship between the length of the historical flow data series with data quality prediction results. By knowing these characteristics, the error correction of analysis results can be expected due to data limitations, so that the Markov Chain Model can be widely applied to optimization of waterworks operations.Results for the Konto River and River showed that the prediction of flow Kwayangan next year with Markov chain models tend to give better results than the results of forecasting by conventional methods are widely applied. Markov model is good enough to predict the river flow has low flow fluctuations, but for a river flow fluctuated sharply less than satisfactory results. The length of data series ranges from 15 to 20 of the optimal inputs to produce a minimum error rate prediction. Accuracy of prediction result is not determined by the length of the input data series, but is determined by the nature of statistical data. Value of lag-1 correlation coefficient are large and small skewness coefficient of the historical data tends to give a satisfactory prediction results.Key words: river flow, data, prediktion, markov model.