Optimasi Parameter Pemulusan Algoritma Brown Menggunakan Metode Golden Section Untuk Prediksi Data Tren Positif dan Negatif

Abstract

Algorithm DES (Double Exponential Smoothing) Brown is a forecasting algorithm used to predict time series data both patterned positive trends and negative trends. However, this algorithm has a weakness in determining the optimum parameter value to minimize forecasting error (MAPE), the parameter value is searched using Golden Section method previously searched manually using repeated experiment. This research uses 60 trend patterned data analyzed for grouping positive and negative trend pattern data which further done forecasting process, evaluation and testing to know what type of data pattern is best. Based on the result, it revealed that optimization parameter yields optimum MAPE value, where parameter value is done forecasting process in positive and negative trend pattern data group yielding average MAPE value equal to 9,73401% (highly accurate) for patterned data positive trend and 15,78467% (good forecast) for negative patterned pattern data. DES Brown forecasting algorithm with parameter optimization method resulted in the approximate value of the original data if the data shows the addition or decrease in value around the average value. Conversely, it will result in a high MAPE value (inaccurate) if the data has a spike in data value periods. From the two groups of MAPE scores, a statistical t test showed that positive trend patterned data (μ1) yielded better MAPE average value than negative trend patterned data (μ2).