The Essentials of Machine Learning in Finance and Accounting

Mohammad Zoynul Abedin, M. Kabir Hassan, Petr Hajek, Mohammed Mohi Uddin, Mohammad Zoynul Abedin, M. Kabir Hassan, Petr Hajek, Mohammed Mohi Uddin

• Publisher: Routledge
• Year: 2021
• ISBN:
  9781000394122
• Edition: 1
• Topic: Business
• Subtopic: Financial Engineering
• Index: Business

This book introduces machine learning in finance and illustrates how we can use computational tools in numerical finance in real-world context. These computational techniques are particularly useful in financial risk management, corporate bankruptcy prediction, stock price prediction, and portfolio management. The book also offers practical and managerial implications of financial and managerial decision support systems and how these systems capture vast amount of financial data.

Business risk and uncertainty are two of the toughest challenges in the financial industry. This book will be a useful guide to the use of machine learning in forecasting, modeling, trading, risk management, economics, credit risk, and portfolio management.

Chapter 1

Machine learning in fnance and accounting

Mohammad Zoynul Abedin, M. Kabir Hassan, Petr Hajek and Mohammed Mohi Uddin

1.1 Introduction

Machine learning (ML) is a type of applied artificial intelligence (AI) that enables computer systems learn from data or observations, and automatically improves predictability by utilizing ongoing learning. It is generally featured in computer science discipline but can be applied in disciplines such as social sciences, finance, accounting and banking, marketing research, operations research, and applied sciences. It utilizes computationally intensive techniques, such as cluster analysis, dimensionality reduction, and support vector analysis. ML has experienced a rise in recognition among academics, researchers, and practitioners over the last couple of decades because of its ability to help predict more accurately. Many applications of ML across diverse fields have emerged. Particularly, its applications in major disciplines such as finance, accounting, information systems, statistics, economics, and operations research are noteworthy. In the context of rapid innovations in computer science and availability of big data, ML can change the way practitioners make predictions, and researchers collect and analyze data.

Computational finance is an interdisciplinary area that integrates computing tools with numerical finance. By utilizing computer algorithms, it can contribute to the advancement of financial data modeling systems. These computational techniques can successfully be utilized in important finance areas, such as financial risk management, corporate bankruptcy prediction, stock price prediction, and portfolio management. For example, ML can be utilized to prevent and detect credit card frauds. Taking into consideration the availability of huge volume of unstructured data, such as customer reviews and social media posts and news data, ML can provide “new insights into business performance, risks and opportunities” (Cockcroft & Russell, 2018, p. 324). ML tools can be utilized to process these unstructured data in order for making better business decisions. Managers can use ML tools in preventing and detecting accounting frauds (see Cockcroft & Russell, 2018). It can also be used in accounting areas such as auditing, income tax, and managerial accounting.

ML can be successfully utilized in accounting and finance research. For example, content analysis is a widely used research method in accounting research. ML algorithms can provide “reliability, stability, reproductivity and accuracy” (see Bogaerd & Aerts, 2011, p. 13414) in data processing. Accounting researchers in corporate social responsibility (CSR) frequently use textual analysis, a sub-category of content analysis, for identifying themes. ML algorithms can successfully be used to classify texts or generate themes (see Bogaerd & Aerts, 2011). Accounting scholars (Deegan & Rankin, 1998; Gray, Kouhy & Lavers, 1995; Neu, Warsame & Pedwell, 1998) used algorithms in classifying texts from corporate social and environmental reports. Bogaerd and Aerts (2011) used LPU (learning from positive and unlabeled data) ML method in classifying texts with 90% accuracy. Behavioral finance researchers can use unstructured newspaper and social media data to understand market sentiment, and utilize the data in developing models predicting prices of financial products.

The next two sections highlight the motivation and provide the brief overviews of chapters appearing in this book.

1.2 Motivation

ML is an emerging computing tool that can be successfully utilized in large and complex data settings. For example, recently researchers from Warwick University, UK, discovered 50 new planets from existing NASA data by using ML algorithms (Yeung, 2020). These types of opportunities were not available in the past due to the absence of large datasets and the processing limitations of computers. Due to the advancement in computing technology, “big data” are now easily available on various business areas. ML can be used to exploit the immense potential of utilizing “big data” for making improved business decisions. Research on utilizing big data (and ML) has potentials for better “industry practices” and “cross-disciplinary research” (see Cockcroft & Russell, 2018, p. 323). However, although it has tremendous potentials, Goes (2014) argues that finance industry does not have sufficient expertise to exploits the benefits of big data. This book is interdisciplinary in nature. It aims to contribute to the emerging machine learning area and its applications in businesses.

This book will present 12 chapters covering topics including machine learning concepts, algorithms and their applications. More specifically, this book introduces methods such as kernel switching ridge regression, sentimental analysis, decision trees, and random forests. It also introduces empirical studies applying ML in multiple finance and accounting areas, such as forecasting of mortality for insurance product pricing, using kernel switching ridge regression for improving prediction models, managing risk and financial crimes, and predicting stock return volatilities. Given the lack of availability of sufficient books in this area, this book will be useful to researchers, including academics and research students, who are interested in advanced machine learning tools and their applications. The contents of this proposed book are also expected to benefit practitioners who are involved in forecasting modeling, stock-trading risk management, bankruptcy prediction, accounting and banking fraud detection, insurance product pricing, credit risk management, and portfolio management. We believe findings from this book will add new insights into the stream of computational finance and accounting research.

1.3 Brief overview of chapters

In addition to this introduction chapter, there are 11 chapters included in the book. We briefly introduce these chapters in the following sections.

Chapter 2 reviews Breiman’s CART algorithm, classification features, and non-parametric methods, i.e., decision trees and random forests (Casarin, Facchinetti, Sorice & Tonellato, 2021). The authors also apply the decision trees and random forests in financial time series in predicting default probability in selected enterprises.

Chapter 3 applies ML to enhance longevity risk management by life insurance companies and pension fund managers. Particularly, this chapter shows how ML can help improve mortality forecasting. The authors used mortality data and the ‘forecasted mortality rates’ in pricing life insurance products (Levantesi, Nigri & Piscopo, 2021).

In Chapter 4, the authors introduced kernel switching ridge regression, an ML method. They argue that the method can make predictions from multiple “regimes of dataset” and “can overcome the unstable solution and the curse of dimensionality” (Alam, Komori & Rahman, 2021, p. 14). By using evidence from an experimental study, the authors show that this method can provide better results than some other popular ML methods.

In Chapter 5, the authors utilized sentiment analysis in predicting stock return volatilities. By analyzing textual and fundamental indicators’ data from annual reports of a large number of US companies, the authors show that ML methods can help generate more accurate predictions of stock price movements (Hajek, Myskova & Olej, 2021).

Chapter 6 introduces some important concepts and ML algorithms, and applications of machine learning techniques in the fields of economics and finance. By reviewing existing literature, the chapter also provides valuable insights to researchers, practitioners, and readers who seek to understand key algorithms used in ML in the field of finance and economics fields. It also examined the effectiveness of ML methods in time series analysis through a simulation study (Casarin & Veggente, 2021).

Chapter 7 focuses on the use of ML and AI in financial services industry. The authors provide examples where ML and AI can transform how the financial services industry can improve products and services, and minimize risk. They used three business cases on “combating financial crimes,” “mitigating risk exposures,” and how “regulators understand potential cloud concentration risk exposures” to justify the potentials of ML and AI in financial services industry (Harmon & Psaltis, 2021, p. 1).

Chapter 8 focuses on the importance of using AI in an audit process. It introduces an AI- based audit framework and explains some benefits and challenges of using AI in an audit process. This chapter introduces the uses of AI by leading audit firms in an audit process with reference to developed countries. This study emphasizes on ensuring transparency in audit process for taking decisions and to giving judgments on various audit affairs.

Chapter 9 introduces web usage analysis that integrates Pillar 3 information assessment in turbulent times. By using data from website visits logs, the chapter assessed the “interests of bank depositors on the requirements of Pillar 3 disclosures and Pillar 3 related information” during the credit crunch-related financial crisis in 2009 (Pilkova, Munk, Blazekova & Benko, 2021, p. 1).

Chapter 10 introduces various ML concepts and algorithms, and applications of ML in accounting, finance, and economics. Particularly, it highlights the importance of using ML in “algorithmic trading and portfolio management, risk management and credit scoring, insurance pricing and detection of accounting and financial fraud” (Radwan, Drissi & Secinaro, 2021, p. 2).

Chapter 11 discusses challenges of applying classification techniques in highly class imbalanced dataset. Select techniques including oversampling, undersampling, SMOTE, and borderline-SMOTE to solve class imbalance problems are presented. This chapter also presents different metrics for evaluating performance of classification techniques applied on imbalanced dataset.

Chapter 12 is about the applications of AI in staff recruitment. This chapter applies the lens of combined system of acceptance and usage of technology. By doing so, it highlights the importance of various antecedents of acceptance of AI among HR experts for hiring talents in Bangladesh. It also identifies the determinants of AI adoptions. By employing a deducting reasoning approach, the authors make some interesting empirical contributions. It also provides some insightful comments and notes on opportunities for further research in this area.

References

1. Alam, M. A., Komori, O., & Rahman, M. F. (2021). Kernel switching ridge regression in business intelligence system. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17—45). Oxford: Taylor and Francis.
2. Bogaerd, M. V., & Aerts, W. (2011). Applying machine learning in accounting research. Expert Systems with Applications, 38, 13414—13424.
3. Casarin, R., Facchinetti, A., Sorice, D., & Tonellato, S. (2021). Decision trees and random forests. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17–45). Oxford: Taylor and Francis.
4. Casarin, R., & Veggente, V. (2021). Random projection methods in economics and finance. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17–45). Oxford: Taylor and Francis.
5. Cockcroft, S., & Russell, M. (2018). Big data opportunities for accounting and finance practice and research. Australian Accounting Review, 86 (28), 1–12.
6. Deegan, C., & Rankin, M. (1996). Do Australian companies report environmental news objectively? An analysis of environmental disclosures by firms prosecuted successfully by the Environmental Protection Authority. Accounting, Auditing and Accountability Journal, 9(2), 50–67.
7. Goes, P. B.(2014). Big Data and IS Research. Minneapolis: Carlson School of Management.
8. Gray, R., Kouhy, R., & Lavers, S. (1995). Corporate social and environmental reporting: A review of the literature and a longitudinal study of UK disclosure. Accounting, Auditing and Accountability Journal, 8(2), 47–77.
9. Hajek, P., Myskova, R., & Olej, V. (2021). Predicting stock return volatility using sentiment analysis of corporate annual reports. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17–45). Oxford: Taylor and Francis.
10. Harmon, R., & Psaltis, A. (2021). The future of cloud computing in financial services: A machine learning and artificial intelligence perspective. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17–45). Oxford: Taylor and Francis.
11. Levantesi, S., Nigri, A., & Piscopo, G. (2021). Improving longevity risk management through machine learning. In M. Z. Abedin, Hassan, M. K., Hajek, P., and Uddin, M. M. (Eds.), The Essentials of Machine Learning in Finance and Accounting (pp. 17–45). Oxford: Taylor and Francis.
12. Neu, D., Warsame, H., & Pedwell, K. (1998). Managing public impressions: Environmental disclosures in annual reports. Accounting, Organizations and Society, 23(3), 265–282.
13. Pi...