Open AccessIMPROVISASI MODEL ARIMAX-ANFIS DENGAN VARIASI KALENDER UNTUK PREDIKSI TOTAL TRANSAKSI NON-TUNAI
Author(s) -
Muhammad Luthfi Setiarno Putera
Publication year - 2020
Publication title -
indonesian journal of statistics and applications
Language(s) - English
Resource type - Journals
ISSN - 2599-0802
DOI - 10.29244/ijsa.v4i2.603
Subject(s) - cash , payment , database transaction , computer science , fuzzy inference system , autoregressive integrated moving average , adaptive neuro fuzzy inference system , econometrics , business , fuzzy logic , mathematics , time series , artificial intelligence , finance , machine learning , fuzzy control system , world wide web , programming language
Developed information technology boosts interest to use non-cash payment media in many areas. Following the high usage of a non-cash scheme in many payment transactions recently, the objective of this work is two-fold that is to predict the total of a non-cash transaction by using various time-series models and to compare the forecasting accuracy of those models. As a country with a mostly dense Moslem population, plenty of economical activities are arguably influenced by the Islamic calendar effect. Therefore the models being compared are ARIMA, ARIMA with Exogenous (ARIMAX), and a hybrid between ARIMAX and Adaptive Neuro-Fuzzy Inference Systems (ANFIS). By taking such calendar variation into account, the result shows that ARIMAX-ANFIS is the best method in predicting non-cash transactions since it produces lower MAPE. It is indicated that non-cash transaction increases significantly ahead of Ied Fitr occurrence and hits the peak in December. It demonstrates that the hybrid model can improve the accuracy performance of prediction.