The Impact of Digital Transformation and FinTech on the Finance Professional

The calculation of key risk indicators (KRI) and key performance indicators (KPI) is widely performed by several subsequent transformation and calculation steps bound together in a batch process. This article introduces certain algorithms (machine learning and deep learning) to derive sources and causes of errors from data and error patterns.

Blockchain technology has recently received a lot of media attention. However, the discussion is often restricted to concepts and not about the technology itself or how to implement it. In this article, we would like to sum up the benefits of blockchain for the financial industry and introduce possible technologies. After providing a brief overview of the various Hyperledger frameworks and tools, we will introduce Hyperledger Fabric. We will be explaining the components of this open-source framework in detail and illustrating a sample transaction process. Special emphasis will be on privacy methods and how they can be implemented using Hyperledger Fabric.

Bitcoin and blockchain are widely discussed subjects among fintech companies and established players in the financial services sector. The nature of this technology can have a disruptive effect on financial services. This article focuses on aspects discussed less frequently in public including digitally agreed algorithms to accelerate processes, smart contracts and the privacy required to ensure only involved parties can see what they need to see. These aspects are illustrated here using a sample implementation of a few selected structures and processes needed for the digitalization of loan syndication using The Linux Foundation’s tool Hyperledger Composer.

This section places the concepts of next best offer (NBO) and next best action (NBA) in the banking context. The author introduces the two concepts (NBO and NBA) with a focus on specific features. It briefly touches on the concept of customer experience and then shows how NBO can be used for forecasting.

The prospect theory model developed by Daniel Kahnemann and Amos Tversky in 1979 is now widely recognized as providing more empirically valid explanations of decision-making under uncertainty than the classical von Neumann-Morgenstern paradigm of expected utility maximization. However, despite compelling potential use cases, industry applications of prospect theory remain scarce. Following an outline of the shortcomings of classical expected utility theory and some key tenets of prospect theory, we discuss how prospect theory can be used in practice to obtain meaningful estimates of investor risk preferences. We then conclude with a case study of the digital investment management firm LIQID and its use of prospect theory as a decision support tool for its clients.

This article describes the application area of predictive analytics within the context of value driver-based planning. We will briefly describe the technical principles and challenges and then turn our focus to aspects of realization based on practical examples. The key to value driver-oriented planning consists of connecting individual value drivers to logical value driver trees, i.e., mapping causal relationships to derive BS or P&L results and balance sheet items at a future point in time. The challenge here is not mapping causal relationships, but identifying value drivers and verifying valid relationships. The use of predictive analytics provides a sustainable and objective foundation for the process of identification and verification, which is illustrated in this article with practical examples including the design of required algorithms using R and the derivation of a future market potential for mortgage loans.

This article introduces a new framework for risk assessment. Risk management is evolving from the well-established one-year horizon dominated by the value-at-risk concept into a multi-period risk projection framework. This enables organizations to compare the planned and actual risk situations, thus ensuring risks taken are in line with the long-term risk roadmap. We also include the principle of value driver analysis in this framework and discuss the potential of agent-based modeling within the context of developing value drivers.

Intraday liquidity risk management has become of increased interest to regulators and financial professionals. Intraday liquidity and the related business processes are characterized by personal connections. Nonetheless, the aggregated view of the participants outside the bank can be clustered and assigned to certain patterns. Risk management always seeks to take an independent opinion in such an environment. This article focuses on how to detect patterns in intraday flows by customer. Additionally, the article shows how these patterns are used to forecast possible customer behavior and how to aggregate the forecast by currency. An improved decision basis is offered by combining possible and correlated client patterns at currency level not only by using machine learning.

Machine learning and artificial intelligence have become increasingly attractive for quantitative risk management in recent years. Today’s significantly enhanced computer power paves the way for the use of more complex models (e.g., artificial networks with a considerable number of nodes), which results in a much higher quality of results. Both the achievable enhanced quality of results as well as an enhanced familiarity with the methodology result in an increasing acceptance of these methods by financial regulators. In this article, we not only address the aspect of reliability of the results obtained using machine learning methods, but also indicate the impact on how credit risk parameters are modeled.

Real estate has always been an important asset class. Massive investment in real estate since the financial crisis 2008 has pushed the demand for integrated and micro-based risk modeling. The data for such risk management is available to banks and other investors, but is not structured and therefore not model-ready. Appraisals are available as printouts as well as pdf files, but the relevant data is only available in semi-structured form and differs by appraisal. A central issue in integrated real estate risk management consists of transferring the relevant information from appraisals into a risk management environment without too much unnecessary human effort. The goal of this article is to describe an approach to identifying and extracting existing semi-structured appraisal information into structured information ready to be consumed by risk models and other applications.

Networks can be observed both within financial institutions and in the interaction between these institutions and the financial industry as a whole. Dealing with these networks and identifying their critical channels presents numerous challenges to strategic, operational and risk decision-makers. Many novel approaches show promise, but have been difficult to scale due to a lack of computing power. Advances in technology however enable financial institutions to implement and use these untapped methods for strategic, operational and risk management. After a brief review of the taxonomy and a presentation of recent methods to study the intrinsic risks, the authors present three distinct modern approaches to dealing with network modeling and their respective management approaches. Neural networks, agent-based modeling and process mining techniques are particularly highlighted here, supported by a range of use cases. Finance professionals are encouraged to review their approaches to dealing with network complexity in the light of recent technological advances in using these methods.

This article section deals with the challenges of a Chief Risk Officer and how to go about solving them. It investigates the potential of big data and an analytics platform including aspects of open source software and the cloud. Challenges are then dealt with in further detail, spanning from financial risk and nonfinancial risk to regulatory aspects and the impact of machine learning, AI and ABM. The section summarizes how the next generation of risk management should look.

An ever-increasing volume of data at financial institutions combined with a regulatory tsunami leads to problems that are difficult to solve with the current methods in order to meet the regulatory requirements. This article investigates how Distributed Ledger Technology (DLT), Artificial Intelligence (AI) and Machine Learning (ML) can open new horizons for the regulatory value chain and can be seen as possible cost reducers. The power of AI and ML is based on the data being used. In this article, we will take a closer look at how DLT could improve regulatory reporting quality and processes and what implications DLT can have for banks and regulators with respect to regulatory reporting. Moreover, it investigates how AI can support us in regulatory reporting, using MiFID 2 as an example. It discusses what risks can be mitigated for Anti Money Laundering (AML), Know Your Customer (KYC) and how the generic intelligent stress tests help to meet the minimum requirements. Lastly, it describes how natural language processing (NLP) can reduce the regulatory burden.

Financial technology (fintech) companies evaluating whether to apply for an OCC Special Purpose National Bank (SPNB) charter must consider a variety of factors. Among them are the regulatory and compliance requirements they will need to adhere to; the work involved in applying for the charter and transitioning to full conformance; the challenges (and opportunities) associated with competing against traditional banks; and the strategic and operational changes that may be required down the road to align with the OCC’s expectations on an ongoing basis.

This chapter classifies the different machine learning algorithms into domains and provides a formal definition of machine learning. In addition, the chapter describes briefly a common set of the classic machine learning techniques. These sets span from time series forecasting to different clustering methods including trees and Bayesian networks. The special domain of deep learning is addressed in the following chapter (Liermann, Li, & Schaudinnus, Deep learning—An introduction, 2019b).

The successful application of deep learning has improved significantly in the last decade. The chapter provides a compact overview of the selected types of neural networks. The deep learning algorithms applied in the use cases in this book are described in more detail here. The chapter concludes with an overview of the common deep learning frameworks and vendor offerings.

Hadoop has become a standard for processing big data in a clustered environment. This article provides an introduction to Hadoop/HDSF and other important Apache projects including Spark, Hive and HBase. The basic concepts like worker nodes and cluster manager are also introduced here.

The concepts of in-memory databases are used in predictive analysis and other high data analysis performance applications. This article explains the basic concepts of in-memory databases as well as their differences and advantages over traditional databases. It primarily includes an in-depth analysis and discussion of the SAP HANA framework, followed by a use case involving a performance optimized Cash Flow Generator.

This section describes the practical use of a document-based database (MongoDB). It first provides a brief history of database environment development. After introducing MongoDB, the section highlights the advantages of the document-based database and describes deployment options.

The book covers a large cross section of the digitalization story with a focus on the financial sector. The introduction highlighted several major digitalization and fintech trends. The second chapter dealt with some of the most interesting aspects such as process automation, distributed ledgers and client-related innovative approaches.