F. Rahutomo, M.M. Huda, R.A. Asmara, A. Setiawan, A.A. Septarina
In foreign exchange money trading, historical data are publicly available continuously. This historical data such as opening, highest, lowest, and closing rate are important variable to predict the future of rates movement. The available data is not only historical trading itself, but also from news release and expert analysis from expert trader. This kind of data contains text and number. This paper proposes in forecasting the rates by combining text and number data. The combination of text mining technique with several time series method i.e: simple moving average, weighted moving average and exponential moving average. Research period for this experiment is between 1st December 2018 and 31st January 2019. The currency pair are EUR - USD, USD-JPY and EUR - JPY. Forecasting results with some time series method were compared with combined time series forecasting method and naïve bayes classifier. The experiment results show that combined time series method with naïve bayes classifier delivered better accuracy level. © Published under licence by IOP Publishing Ltd.
Information Technology Department, State Polytechnic of Malang, Jl. Soekarno-Hatta No 9, Malang, 65141, Indonesia; Electrical Engineering Departement, State Polytechnic of Malang, Jl. Soekarno-Hatta No 9, Malang, 65141, Indonesia