Machine Learning and Network Analytics in Finance
Permanent URI for this collection
1 - 5 of 5
ItemEvent Entry Time Prediction in Financial Business Processes Using Machine Learning - A Use Case From Loan Applications( 2018-01-03)The recent financial crisis has forced politics to overthink regulatory structures and compliance mechanisms for the financial industry. Faced with these new challenges the financial industry in turn has to reevaluate their risk assessment mechanisms. While approaches to assess financial risks, have been widely addressed, the compliance of the underlying business processes is also crucial to ensure an end-to-end traceability of the given business events. This paper presents a novel approach to predict entry times and other key performance indicators of such events in a business process. A loan application process is used as a data example to evaluate the chosen feature modellings and algorithms.
ItemA Partial Parameter HMM Based Clustering on Loan Repayment Data: Insights into Financial Behavior and Intent to Repay( 2018-01-03)Financial institutions that provide loans are interested in understanding, as opposed to just predicting, the repayment behavior of its customers. This study applies a modified Hidden Markov Model (HMM) based clustering which clusters repayment sequences across selected subsets of the HMM parameters. We demonstrate that different implementations of this adaptation help us gain an in-depth understanding of various drivers that are hard to observe directly, but nevertheless govern repayment. These include drivers such as the ability to repay, or the intention to repay independent of the ability. Our results are compared to an alternate sequence clustering approach. The study concludes with the observation that the ability to cluster on selective parameters, in conjunction with the structural construct of HMMs, enables the discovery of substantially more meaningful business insights.
ItemCredit Risk Evaluation in Peer-to-peer Lending With Linguistic Data Transformation and Supervised Learning( 2018-01-03)The widespread availability of various peer-to-peer lending solutions is rapidly changing the landscape of ï¬ nancial services. Beside the natural advantages over traditional services,a relevant problem in the domain is to correctly assess the risk associated with borrowers. In contrast to traditional ï¬ nancial services industries, in peer-to-peer lending the unsecured nature of loans as well as the relative novelty of the platforms make the assessment of risk a difï¬ cult problem. In this article we propose to use traditional machine learning methods enhanced with fuzzy set theory based transformation of data to improve the quality of identifying loans with high likelihood of default. We assess the proposed approach on a real-life dataset from one of the largest peer-to-peer platforms in Europe. The results demonstrate that (i) traditional classiï¬ cation algorithms show good performance in classifying borrowers, and (ii) their performance can be improved using linguistic data transformation
ItemRiskrank to Predict Systemic Banking Crises With Common Exposures( 2018-01-03)Systemic risk has remained at the nexus of macro-financial research and policymaking in most parts of the world. Much of the attention has focused on understanding implication of the interconnectedness of financial markets. Instead of focusing only on networks, we use and test the utility of network structures in a novel way. We use RiskRank as a framework to test the use of networks of financial systems, and particularly focus on testing the utility of the network dimension of common exposures (funding composition and portfolio overlap). RiskRank provides an ideal playground for testing the extent to which direct and common exposures perform in capturing transmission of financial crises. The results in this paper highlight the importance of common exposures. We show that funding and portfolio composition overlap are significant channels of contagion and should be accounted for when measuring systemic risk.