VIETNAM NATIONAL UNIVERSITY, HANOI INTERNATIONAL SCHOOL GRADUATION PROJECT MACHINE LEARNING MODELS TO PREDICT SHAREHOLDER RETURNS IN THE AIRLINE INDUSTRY Vũ Ngô Bảo Châu Hanoi – Year 2024 VIETNAM NATIONAL UNIVERSITY, HANOI INTERNATIONAL SCHOOL GRADUATION PROJECT MACHINE LEARNING MODELS TO PREDICT SHAREHOLDER RETURNS IN THE AIRLINE INDUSTRY SUPERVISOR: DR. LÊ ĐỨC THỊNH STUDENT: VŨ NGÔ BẢO CHÂU CODE: 20070904 COHORT: BDA2020B MAJOR: BUSINESS DATA ANALYTICS Hanoi – Year 2024 ACKNOWLEDGEMENT I extend my heartfelt thanks to my thesis advisor, Dr. Le Duc Thinh from the International School at Vietnam National University, whose steadfast support and expert guidance have been instrumental throughout my academic endeavors. His patience, encouragement, and profound knowledge have significantly enriched my research experience and thesis development.
I am immensely grateful for having such an exceptional mentor during my academic tenure at the university. Furthermore, I would like to express my genuine appreciation to my friends for their invaluable support throughout my studies at VNU-IS and the thesis writing period. Their unwavering support has not only contributed to my academic achievements but has also played a crucial role in my personal development. I am deeply thankful for their presence and encouragement.
1 LETTER OF DECLARATION I hereby declare that the Graduation Project titled "Machine Learning Models to Predict Shareholder Returns in the Airline Industry" is the product of my independent research and has not been published in other works. During this project, I have rigorously adhered to research ethics. All findings presented are the result of my research and surveys. Furthermore, all references used in this project have been properly cited by academic standards.
I take full responsibility for the accuracy of the data, numbers, and all other content included in my graduation project. Hanoi, June 16th 2024 Vũ Ngô Bảo Châu 2 ABSTRACT The research aims to understand how the use of different machine learning approaches in predicting shareholder returns to identify the efficiency of airline companies as this sector plays a critical role in connecting countries and generating positive effects on their economies. However, the industry has significant heft, and airlines often fail to consistently provide shareholder value, making it crucial to improve the current predictive modeling. This paper employs Random Forest, AdaBoost, XGBoost,… as analytical models to study the financial data of nine leading U.
airlines over the period 2007 to 2023. Based on this metric fixation on Total Shareholder Return (TSR), this paper examines which machine learning models are capable of the best financial prediction. Conclusions suggest that compared with each predictor, ensemble methods perform better and provide deep and valuable information regarding investment decisions. This paper proves that the data reveal that improvements in predictive accuracy and Shareholder value in a volatile airline industry can be achieved through machine learning hence this supports the stakeholders to adopt strategic financial planning.
3 LIST OF ABBREVIATION ANA All Nippon Airways ASK Available seat kilometers ASM Available seat miles BMA Bayesian Model Average COGS Cost of goods sold CASM Cost per available seat miles D/E Debt to Equity EPS Earning per share IAG International Consolidated Airlines Group PRASM Passenger revenue per available seat miles RASM Revenue per available seat miles ROA Return on Assets RPK Revenue passenger kilometers RPM Revenue passenger miles TSR Total stock return 4 LIST OF FIGURE AND TABLE Figure 3. A part of the final dataset used for building prediction models. Correlation analysis table. The summary value of all data.
The descriptive analysis table. Result summary for Random Forest Regression model. Actual versus Predicted Result for Random Forrest model. Adaboost implementation method.
Summary result for AdaBoost model. Actual versus Predicted Result for AdaBoost model. Summary result for XGBoost model. Actual versus Predicted Result for XGBoost model.
Artifical Neuron Networks Layer. Summary result for ANN model. Actual versus Predicted Result for ANN model. Model fitted using Linear kernel.
Summary result for SVR model. Actual versus predicted result SVR model. Decision Tree Model. Summary result for Decision Tree model.
Actual versus Predicted Result for Decision Tree model. Results from machine learning models for TSR. 28 5 TABLE OF CONTENTS ACKNOWLEDGEMENT. 1 LETTER OF DECLARATION.
3 LIST OF ABBREVIATION. 4 LIST OF FIGURE AND TABLE. 5 TABLE OF CONTENTS. The necessity of topic.
Total Shareholder Return (TSR). The objective of topic. 10 CHAPTER 2: LITERATURE REVIEW AND RESEARCH METHODOLOGY. Research problem, methodology and scope.
Scope of research. Research Challenges and Resources. 18 6 CHAPTER 3: MAIN RESULTS. Performance Evaluation Metrics.
Machine Learning for Prediction Models. 28 CHAPTER 4: CONCLUSION, IMPLICATION AND RECOMMENDATION. Recommendations for investors. Limitation of study & recommendations for further research.
Recommendations for future research. The necessity of topic 1. Context The aviation industry serves as a pivotal element in global connectivity, driving economic expansion and enhancing societal mobility. However, despite its significant role, airlines have often faced difficulties in consistently delivering strong shareholder returns.
Tony Tyler's report "Profitability and the Air Transport Value Chain" from IATA emphasizes the stark contrast between the rapid expansion of air travel and cargo and the relatively modest profitability of airlines (Tony Tyler, 2013). This highlights the ongoing challenges in managing shareholder returns effectively within the aviation sector. In today’s data-rich environment, where data-driven strategies are crucial, machine learning offers valuable opportunities for improving decision-making and enhancing predictive analytics. The aviation industry, characterized by its intricate operations and significant risks, is particularly well-positioned to benefit from the deployment of these advanced models.
Research such as the 2013 executive compensation study by the State Board of Administration (SBA) illuminates the importance of aligning performance metrics with shareholder value (State Board of Administration, 2013). Yet, the selection of these metrics often requires customization to fit the unique demands of different industries, including aviation. Despite extensive research on the application of machine learning in areas like ticket pricing and flight delay prediction, there is a noticeable research gap regarding the use of these technologies to predict shareholder returns in the aviation sector. This area presents a significant opportunity for developing more focused and industry-specific machine-learning applications.
Total Shareholder Return (TSR) Total Shareholder Return (TSR) is an essential metric used to evaluate the performance of investments in the airline industry over a given period. It includes capital appreciation, reflected by changes in stock prices, and dividends paid to shareholders, offering a comprehensive view of the returns generated from owning a company's stock (Brigham & Ehrhardt, 2013). TSR is particularly critical in the airline industry due to the sector's volatility and the significant impact of external factors such as fuel prices and economic conditions on stock performance. In this study, TSR is a critical measure for assessing the effectiveness of machine learning models in forecasting shareholder returns in the airline sector.
This research leverages TSR data from trusted financial sources like Yahoo Finance (Yahoo Finance, 2024) to explore the predictive strength of these models and their potential impact on investment strategies. The use of reliable financial data ensures that the analysis is grounded in accurate and up-to-date information, which is essential for developing robust predictive models. TSR is a vital indicator for investors as it encapsulates the combined effects of stock price movements and dividend yields. A positive TSR denotes value creation, indicating a favorable increase in investment returns.
Conversely, a negative TSR points to a reduction in shareholder value, which may signal underlying issues or inefficiencies within the airline industry. By examining TSR, investors can gain insights into both the short-term and long-term performance of their investments, helping them to make more informed decisions. By utilizing TSR data from platforms such as Yahoo Finance, researchers can tap into a rich repository of financial information crucial for a thorough analysis (Yahoo Finance, 2024). This study aims to deepen the understanding of the variables that influence shareholder returns, shedding light on the complex dynamics of the airline industry through the lens of advanced predictive modeling.
The integration of machine learning techniques such as regression analysis, decision trees, and neural networks will enable a comprehensive evaluation of the factors 9 driving TSR, ultimately enhancing the predictive accuracy and utility of financial forecasts in the airline sector. The objective of topic This thesis aims to rigorously investigate how specific financial ratios and other pertinent metrics influence Total Shareholder Return (TSR) in the airline industry. The study collects yearly financial data from the world's top airlines to calculate these metrics and analyze their correlation with TSR. This data serves as the foundation for developing a comprehensive multiple regression model that assesses the impact of each metric on TSR.
Research outcomes The study meticulously analyzes data from nine prominent U. airlines, focusing on Total Shareholder Returns (TSR) and a combination of specific airline industry metrics alongside conventional corporate finance metrics. By calculating correlations and employing a multiple linear regression model, this research aims to uncover potential linear relationships positive and negative between these metrics and TSR, providing insights into how these relationships influence stock returns. Practical contributions This research delineates its practical contributions by offering actionable insights that can assist airline companies, especially within the Vietnamese context, in refining their financial and operational strategy disclosure.
These insights are aimed at enhancing investment attraction by showcasing how critical metrics influence TSR, thereby helping airlines to strategically manage and report their financial performance to maximize shareholder value. 10 CHAPTER 2: LITERATURE REVIEW AND RESEARCH METHODOLOGY 2. Theoretical background The airline industry's performance is deeply influenced by a variety of financial and operational metrics, as highlighted in the diverse sources referenced. For instance, the study sponsored by the State Board of Administration (SBA) and conducted by Farient Advisors LLC illustrates how executive compensation is aligned with shareholder value, emphasizing metrics like earnings growth, returns, and revenue growth, which are found to have significant impacts on stock prices (Farient Advisors LLC, 2011).
This aligns with economic theories that suggest well-aligned performance metrics contribute to enhanced shareholder returns. Further, discussions by industry experts such as Gary Leff and Ted Reed highlight the market's deep focus on unit revenue metrics like PRASM and RASM, despite the apparent disconnect between these metrics and the actual stock performance as noted by industry executives (Leff, 2016; Reed, 2016). This suggests a complex interaction between reported financial metrics and market perceptions that influence stock prices. To navigate the complexities of financial data, the adoption of machine learning models presents significant advantages.
By employing Artificial Neural Networks (ANNs), Random Forest, and Support Vector Regression (SVR), these models excel at uncovering the nonlinear relationships and subtle intricacies in data that traditional economic models often overlook. These sophisticated tools are adept at processing large datasets, enabling the identification of patterns that might elude human analysts, thus enhancing the accuracy of forecasts for shareholder returns. Additionally, the integration of these models is underpinned by core economic theories such as the Efficient Market Hypothesis (EMH) and the Arbitrage Pricing Theory (APT). EMH posits that stock prices incorporate all available information, implying that achieving returns above the market average is difficult without the 11 deeper insights provided by machine learning models (Fama, 1965).
Conversely, APT provides a multifactorial framework for analyzing stock returns, aligning well with the comprehensive capabilities of machine learning technologies to assess numerous variables simultaneously (Ross, 1976). These theoretical models not only support but also enhance the utility of machine learning in financial analysis. In conclusion, the theoretical background for using machine learning to predict shareholder returns in the airline industry combines these economic theories with practical insights from industry studies and expert commentary. By leveraging advanced analytics and comprehensive data analysis, stakeholders can gain a deeper understanding of the factors driving stock performance in the airline sector, aiding in more informed decision-making for long-term investment strategies.
Total Shareholders Return (TSR) TSR is defined as the total return a stock provides to an investor, encompassing both capital gains and dividends received.