FOREX PREDICTION SYSTEM DEVELOPMENT USING RANDOM FOREST, XGBOOST, AND EXTRA TREES.

• Main Author: Dwinanda Putra, Rahma
• Format: Final Project
• Language: Indonesia
• Subjects: Teknik (Rekayasa, enjinering dan kegiatan berkaitan)
• Online Access: https://digilib.itb.ac.id/gdl/view/70235
• Institution: Institut Teknologi Bandung

Forex market is the largest financial market in the world. The transaction volume in the forex market can reach trillions of USD per day. This fact entices people to trade in the forex market and become a trader. The transaction volume causes the forex market to move volatile so that traders can earn and/or lose up to thousands of USD only in a couple of days. The research aims to design a prediction model that can be used to predict the forex market using data mining. The prediction model is designed based on the technical analysis that is commonly used by traders, such as indicator analysis, time series analysis, and pattern recognition. Prediction model is built using 3 data mining algorithms: there are random forests, XGBoost, and extra trees. The three models will be compared with each other to obtain the best model with the highest accuracy that will be used for prediction. The prediction model that is designed is on the fly where the prediction model will always change during the system run. This is aimed so that the resulting prediction model can adjust to the forex market which is always updated at any time. The research resulted in a forex prediction model and prototype of a prediction system. Users only need to input a pair of currency that will be predicted and minimum profit in pips that they want to earn. Pips is unit of change that occur in forex. The output of the prototype of prediction system is a transaction decision suggestion in the forex market in the form of buy, sell, or do nothing with the accuracy rates exceeding 80%.