| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系 === 107 === With the foreign exchange market located around the world, the overlapping of working hours in different time zones allows the foreign exchange to reach a 24-hour non-stop transaction on weekdays. There are many types of transactions in the foreign exchange market. Such as spot and out forward transaction, as well as different trading methods for futures and margins, which have contributed to the huge transaction amount. Therefore, foreign exchange trading needs to be improved with the help of information systems. In view of the importance attached to data science in recent years, various machine learning analysis tools have attracted attention. In this thesis, we use the Decision Tree to make three kinds of forecast classes for forex data, which are Up, Down and Keep, to provide investors with an assessment and reference for trading.
In this thesis, we use the Yahoo Financial website to climb price information such as the opening and closing of forex, and simultaneously calculate relevant technical indicators, such as Relative Strength Index (RSI), Stochastic Oscillator (STC), and Moving Average (MA). After analyzing its characteristics and performing data preprocessing, input it into the decision tree for training to find the best prediction results. The virtual trading system of this thesis is to make virtual trading of foreign exchange through the result of forecasting, and directly evaluate the performance of this system by the rate of return. The maximum rate of return at M1 time interval reached 3% in half a month, the H1 time interval reached 6% in 7 months, and the D1 time interval reached nearly 80% in 6 years.
|
|---|
