PERBANDINGAN ALGORITMA GENETIKA DAN METODE STATISTIK ARIMA UNTUK PREDIKSI DATA TIME SERIES (Studi Kasus : Kunjungan Museum di Kota Yogyakarta)

Abstract

Prediction of time series aims to control or recognize the behavior of the system based on the data in the time period, using statistical methods approach, time series prediction techniques were also developed on the model of artificial intelligence. This study aims to create applications by comparing and looking for models that have better performance in making predictions. Prediction using ARIMA taken through three stages, namely: identification of ARIMA models, estimation of the parameters of the model have been identified, and the evaluation of the suitability of the model that has been estimated and predictive ability. Performance comparation is seen from the MSE and MAPE.The empirical studies have shown MSE generated ARIMA model showed minimal error of 0.000198722 and models AG amounted to 28.2153. MAPE percentage value resulting prediction accuracy of the model AG 65.78% and 18.11% of the ARIMA model.
Performance of MSE and MAPE value shows that the model prediction time series AG has better prediction accuracy than ARIMA as evidenced from the value of MAPE (Mean Absolute Percentage Error), but for the smallest MSE value on ARIMA models.

Keywords: time series, ARIMA, genetic algorithm