MINISTRY OF EDUCATION AND TRAINING THE STATE BANK OF VIETNAM HO CHI MINH UNIVERSITY OF BANKING GRADUATION THESIS BUILDING A LOSS GIVEN DEFAULT PREDICTION MODEL BY MACHINE LEARNING TECHNIQUES MAJOR: FINANCE - BANKING CODE : 7340201 TRAN NHAT NAM HO CHI MINH CITY, 2024 MINISTRY OF EDUCATION AND TRAINING THE STATE BANK OF VIETNAM HO CHI MINH UNIVERSITY OF BANKING GRADUATION THESIS BUILDING A LOSS GIVEN DEFAULT PREDICTION MODEL BY MACHINE LEARNING TECHNIQUES MAJOR: FINANCE - BANKING CODE : 7340201 Student name: TRAN NHAT NAM Student code: 050607190277 Class: HQ7-GE14 ADVISOR DR. NGUYEN MINH NHAT HO CHI MINH CITY, 2024 i ABSTRACT The process of Vietnam's economy becoming more integrated with the world economy has made financial activities in our nation easier. The banking industry has changed as a result of this integration, indicating a new stage in the growth of Vietnam's banking system in terms of both quantity and quality. However, banking risks are unavoidable and have the potential to result in large financial losses, endangering the stability of the banking system, the financial market, and the Vietnamese economy as a whole due to the sensitive nature of the banking sector and its susceptibility to various factors.
The growing number of businesses declaring bankruptcy highlights how urgent it is to determine the probability of default in different industries. Given this urgency, the objectives of this study are to: (i) explore models for estimating enterprises' loss given default in Vietnam's commercial banking system using suitable machine learning techniques; (ii) determine the most appropriate model for estimating loss given default of enterprises in Vietnam's commercial banking system through machine learning methods; and (iii) select a suitable model to estimate default probabilities of enterprises in Vietnam's commercial banking system using machine learning, The source of information that collected comes from financial reports of businesses that are borrowing capital from Vietnamese commercial banks and businesses that are operating and listed on Vietnam's financial market. Financial reports of operating businesses were collected in the period from 2009 to 2020. Information about businesses is encrypted to ensure information security.
ii DECLARATION I declare that this thesis has been my own work and that it has not been submitted, in whole or in part, in any previous application for a degree. Unless otherwise stated by reference or acknowledgement, the work presented is entirely my own. The author Tran Nhat Nam iii ACKNOWLEDGEMENTS I express my sincere gratitude to the Board of Directors of Ho Chi Minh University of Banking. My formal university-level training program, the High Quality Program with a focus on Finance - Banking at the school, was organized and made possible by Ho Chi Minh City.
Simultaneously, I express my heartfelt gratitude to all the educators who took part in the program, sharing their knowledge with me and the other students, and overseeing the departments of the school while I was there. We hope the educators stay well and continue to shape the next generation of leaders in the most efficient manner. I would especially like to thank Dr. Nguyen Minh Nhat, who helped me immensely throughout the research process and directly guided me in order for me to successfully finish this thesis.
He also imparted invaluable knowledge and experience. Lastly, I would like to express my gratitude to my family, all of my HQ7-GE14 classmates, and my friends for their unwavering support and encouragement during my trying times. The learning environment that has been established by the school's research and study processes is ideal for me to finish this graduation thesis. iv TABLE OF CONTENT ABSTRACT.
III LIST OF ABBREVIATIONS. VI LIST OF FIGURES. VII LIST OF TABLE. VIII CHAPTER 1: INTRODUCTION.
THE URGENCY OF THE RESEARCH. RESEARCH SUBJECT AND SCOPE. THE STRUCTURE OF RESEARCH. 9 CHAPTER 2: LITERATURE REVIEW.
LOSS GIVEN DEFAULT. OVERVIEW OF THE MODELS USED TO PREDICT THE LOSS GIVEN DEFAULT OF CORPORATE. 18 CHAPTER 3: DATA AND METHOD OF RESEARCH. RESEARCH MODELS AND METHODS.
PYTHON IS USED IN MODEL BUILDING. 39 CHAPTER 4: RESEARCH RESULT AND DISCUSSION. CRITERIA FOR MODEL VERIFICATION AND EVALUATION. 52 CHAPTER 5 : CONCLUSION AND RECOMMENDATIONS.
TOPIC LIMITATION AND POTENTIAL RESEARCH DIRECTIONS. Potential research directions. 70 vi LIST OF ABBREVIATIONS CRE Cash Return on Equity GBM Gradient Boosting Machines LGD Loss Given Default MAE Mean Absolute Error MAPE Mean Absolute Percentage Error MSE Mean Squared Error OLS Ordinary Least Square RF Random Forest SMEs Small and Medium Enterprises RMSE Root Mean Squared Error XGBoost Extreme Gradient Boosting vii LIST OF FIGURES Figure 3.1: Steps to build a machine learning model .2: The figure on the left partitions the space using the binary decomposition method. The figure on the right shows the tree corresponding to the partition on the left .3: The structure of the decision tree .4: Implementation process of random forest algorithm .1: Proportion of default companies .2: Describe the correlation of variables in the estimated model .3: Estimation results of the models in lgd estimation .4: Estimation results of decision trees in lgd estimation.
49 viii LIST OF TABLE Table 2.1: Four types of financial ratios .1: Predictor variables used in the model .2: Table describing the predictor variables in the model .3: Descriptive information about the dependent variables in the estimated model .4: LGD prediction results of models on out-of-sample data sets. 50 1 CHAPTER 1: INTRODUCTION In this chapter, the author introduces an overview of the research, reasons for choosing the topic, research objectives, research questions, objects and scope of the research, research methods, and main content of the thesis. RESEARCH BACKGROUND First, the COVID-19 pandemic appeared in 2019 and developed extremely complicatedly and led to many serious impacts. The global economy is now recovering and gradually reopening, and vaccination campaigns are continuing.
But businesses around the world are more or less "ruined" by COVID-19 when the pandemic has not ended yet. Not only that, but with the appearance of the Delta variant, or more recently, the Omicron variant, which has the ability to spread quickly and strongly, the fight against COVID-19 in the world has entered a more difficult phase. The wave of bankruptcy of large companies and businesses around the world was formed in mid- 2020, when the COVID-19 pandemic affected most major countries in the world. Second, the COVID-19 epidemic has also devastated Vietnamese businesses since 2019.
According to the World Bank's report "Impact of COVID-19 on Businesses in Vietnam: A Rapid Survey on Businesses and COVID-19" (2020), about 50% of small businesses and more than 40% of small businesses Medium businesses have to close temporarily or permanently due to the impact of the COVID-19 epidemic. In 2021, the number of businesses temporarily suspending business for a certain period of time is nearly 55,000, an increase of 18% year-on-year; 48.1 thousand businesses temporarily suspended operations waiting for dissolution procedures, an increase of 27.7 thousand enterprises completed dissolution procedures, down 4.8 thousand enterprises with capital of less than 10 billion VND, down 4%. There are 211 enterprises with capital of over 100 billion VND, down 20. On average, each month, about 10,000 companies leave the market because they cannot "live with" the intensity of the COVID-19 pandemic.
2 Third, with the economy being heavily affected by the COVID-19 pandemic and increasing corporate bankruptcies, the internal credit rating system of commercial banks plays an extremely important role. important in assessing risk when lending to customers. In this way, it not only helps banks make accurate and effective lending decisions but also contributes to effective risk management. Commercial banks in Vietnam are increasingly aware of the importance of this credit rating system for their lending and risk management activities.
In particular, in the context of commercial banks trying to comply with Basel II standards, applying and developing an internal credit rating system has become more necessary than ever. This enhances the ability of credit institutions to measure, evaluate, and manage risk while also helping to improve the reliability and strength of the banking system in the face of financial and business challenges. THE URGENCY OF THE RESEARCH In the context of facing special economic and financial challenges due to the impact of the pandemic and volatile market conditions, the urgency of research on credit rating models becomes increasingly more important and clearer. First, current credit rating models are still not perfect and face some significant limitations.
The diversity and disagreement about the reliability of credit rating models have created a challenge for researchers and risk managers, making choosing an appropriate model difficult. more difficult than ever. Credit ratings play an undeniable role in assessing and estimating the default probability of companies (Huseyin & Bora, 2009). The research work of Aysegul Iscanoglu (2005) and Hayden & Daniel (2010) analyzed a series of credit rating models, including discriminant analysis models, logit models (logistic regression), decision trees, artificial neural networks (ANN), probit regression models, and many other models, thereby highlighting the advantages and disadvantages of each model.
Detailed studies of these models have been conducted, 3 such as Platt (1991) using a logit model to examine and select financial variables and suggesting that standard financial variables from industry should be used in bankruptcy reports instead of just company financial variables. Lawrence (1992) uses a logit model to estimate the probability of mortgage default. Altman (1968) used a differential analysis model. This shows that research and development of credit rating models is a potential research area and has a far-reaching impact on risk management and the sustainable development of the financial global system.
Second, identifying financial indicators that influence the outcome of credit ratings has always been an important goal and research problem in the field of loss given default. From 1926 to 1936, researchers used basic financial indicators for rankings and proposed various methods. For example, Ramser & Foster (1931) used an equity/total net income index, while Fitzpatrick (1932) used an equity/fixed assets ratio. In the next stage, Altman (1968) used a series of financial indicators in different analytical models to predict the likelihood of corporate bankruptcy, including owner’equity/book value, corporate net income/total assets, operating income/total assets, profit after tax/total assets and working capital/total assets.
However, Deakin (1972) used a different method, selecting 14 different financial variables to assess the possibility of bankruptcy, including indicators such as cash/current debt, real cash flow/total debt, and working capital/net sales. As time has passed, researchers have discovered many new financial indicators that can influence credit rating results. For example, Blum (1974) used market returns and quick ratios in his models. Meanwhile, Back, Laitinen, Sere & Wesel (1996) used 31 different indicators in their study.
Third, the credit rating method at commercial banks in Vietnam today is often subjective, qualitative, and based on the assessment experience of credit officers directly serving customers (called professional solutions). This leads to there being no reliable scientific basis for estimating the possibility of bankruptcy of a business, which only 4 supports lending decisions and is not the basis for making decisions. Up to now, in Vietnam, there have been very few published studies on choosing a model to estimate the probability of corporate default based on financial indicators. This creates a need for continued research and development of scientific and effective credit assessment and credit risk management methods in the Vietnamese market.
Fourth, the government has introduced supportive measures and created a favorable legal environment so that the credit rating department can operate effectively and enhance the transparency of financial information. This not only helps banks manage losses from default at an early stage but also supports the stock and bond markets. Researching and selecting appropriate rating models will make an important contribution to the development of credit rating activities in Vietnam. In this context, risk management measures and enhancing transparency in the financial sector play an important role.
The government issued Decree No.