Applying Machine Learning to forecast Taiwan Government 10-year Bond Yield

• Main Authors: LI-MEI YUAN, 袁麗梅
• Other Authors: Chun-Nan Chen
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/764j54

碩士 === 國立臺灣科技大學 === 財務金融研究所 === 106 === The purpose of this study is to use machine learning to predict Taiwan’s 10-year Government Bond yield, using Macroeconomic Indices and information of major financial markets as input data to build predictive models. The predictive model can assist bond traders in making trading decisions through the use of powerful algorithms to spot patterns in the data and forecast bond yield, and predict future interest rate rise or fall at a certain point in time. In recent years, random forest algorithm has been widely used in the field of data science and machine learning, and has high performance in processing classification and regression. In this study, the Python machine learning library Random Forest is used to build a regression model to predict the yield, and it also implements a classification model to predict the upward or downward movement of interest rates. There are three main empirical conclusions. First, the accuracy of the regression model is high, but when the market interest rate occasionally deviates from the inertia interval, there is a great difference between the predicted value and the actual value of the model. Second, the US Government Bond yields are important features in predicting Taiwan Government 10-year bond yield. Third, the accuracy of using this classification model to predict the interest direction of the 20-day rate is as high as 90%. This study has gone through the process of data collection, data cleaning, feature selection, training model and model building. The resulting model has high predictive accuracy, and can provide a reference for traders to analyze market as a trading strategy. However, the data of the input features of this study still substantial room for improvement, the model efficiency can be adjusted, or other algorithms can be used to enhance the predictive efficiency. This study hopes that traders can have the ability to scientifically analyze the financial data in order to obtain good trading performance, and thus attract more trading participants to activate the Taiwan’s government bond market.