Master of Science (MSc)
Mathematics for Science and Finance
Faculty of Science
Xu (Sunny) Wang
After the financial crisis in 2008, for many companies, their credit ratings were downgraded to the non-investment grade. People started concerning the reliability of credit ratings. Credit score rating plays a vital role in the financial system by balancing information between investors and creditors. It is considered as an essential factor to make financial investment decisions. This thesis is an attempt to determine how to predict the credit rating using the publicly available financial information about companies. The data collected are viewed as high-dimensional multivariate financial time series data, which have more than one time series and more than one variable to consider. In our research, the Dynamic Time Warping (DTW) is used to convert the information contained in the high dimensional time series data into a similarity or dissimilarity high-dimensional matrix. Then, the Principal Component Analysis (PCA) is used to perform dimension reduction and extract the important information from the similarity or dissimilarity matrix generated by DTW. Finally, we employ a statistical learning method, namely, the Decision Tree (DT) to predict credit ratings. Furthermore four different scaling methods and several strategies of increasing the sample size have been considered to improve the prediction accuracy. The most encouraging result is that the predicted credit ratings in test data have on average at most a two-grade difference with the true credit ratings.
Ma, Yinduo, "Predicting Credit Ratings with Statistical Learning Methods" (2018). Theses and Dissertations (Comprehensive). 2109.