SCRIBE: Semantic Credit Risk Assessment of Business Ecosystems

Lead Research Organisation: Royal Holloway, University of London
Department Name: Sch of Management

Abstract

This proposal addresses the Digital Economy and Financial Services research challenge by improving Small and Medium Enterprises' (SMEs) access to credit. The issue is that information in and around credit decision-making is generally limited to company and individual track record. It ignores the position and importance of a company in its business ecosystem. Credit lending decisions by finance providers therefore have unseen network effects and limit growth in unseen ways.

To address this issue, SCRIBE uses emerging semantic technologies to provide disruptive innovation in the form of more accurate real-time credit risk assessment based on a dynamic understanding of the position and value of a company in relation to its business ecosystem (or network). The scientific contributions of SCRIBE are twofold. First, the project fuses the state-of-the-art in (social) network analytics and credit assessment techniques to develop its ecosystem-based understanding (and associated marketing opportunities). Second, as technical foundation, the project develops a state-of-the-art method to 'harmonise' the different conceptual models that underlie data drawn from multiple sources, preserving contextual richness in so doing. Contextual preservation is important not only for network-based decision-making, but also for audit and the legal issues considered by the project since it is relatively well-acknowledged that conventional data modelling implicitly abstracts away important aspects of context.

The scientific contributions are developed and exploited via a collaborative partnership that combines understanding of credit risk and assessment at both the transaction-level (via open online accounting data and via collaboration with Lloyds) and firmographic-level (via collaboration with Creditsafe). Addressing the NEMODE ethos, the project maintains a focus on impact via the development of novel information products and applications (via collaboration with Level Business).

Planned Impact

The scale of impact goes from individual companies, to networks of companies (eco-systems), networks of ecosystems, the National Economy and policy-making. In the short-term (and the 3-year duration of the project) we expect that our research will impact initially on the commercial partners involved. We foresee that the initial experimentation of the Information Product, the network-based credit model and the semantic integration hub will influence our commercial partners' businesses (e.g., products/services designed around more accurate and integrated data, credit decisions based on eco-system models, etc.). In the medium term (5-10 years) the impact will be on financial institutions and credit rating agencies, specifically in the way businesses are assessed and money is loaned. Over the same timeframe we foresee the economic model proposed here for credit to inspire other researchers who will adopt this theory to explore its application to other economic problems. In the long term (10-25 years) we envision that the deliverables of this project will have affected and rippled throughout the U.K. Economy and affect government in their policy-making by basing their economic policies and modelling also in the network effects that business eco-systems produce.

More traditionally, we will utilise academic outlets (conferences and journals) - particularly those where there is industry crossover such as the Semantic Technology Conference (SemTech), International Data Protection and Privacy Commissioners' conference and CEE Credit Risk Management. We will engage actively within the EPSRC NEMODE Network in order to present our research and to seek collaboration and support with other investigators funded by the programme. Engagement here will include: (a) organisation of and participation in NEMODE community meetings; (b) workshops at Brunel University to which NEMODE researchers will be invited to present research outcomes and discuss future progress and collaborations.

Publications

10 25 50
publication icon
De Cesare S. (2016) 4D-SETL a semantic data integration framework in ICEIS 2016 - Proceedings of the 18th International Conference on Enterprise Information Systems

publication icon
Tosetti E (2019) A computationally efficient correlated mixed probit model for credit risk inference in Journal of the Royal Statistical Society: Series C (Applied Statistics)

publication icon
Vinciotti V (2019) The effect of interfirm financial transactions on the credit risk of small and medium-sized enterprises in Journal of the Royal Statistical Society: Series A (Statistics in Society)

Related Projects

Project Reference Relationship Related To Start End Award Value
EP/L021250/1 30/06/2014 27/11/2016 £676,567
EP/L021250/2 Transfer EP/L021250/1 28/11/2016 31/01/2018 £194,648
 
Description The project developed the following. First, a way to semi-automatically extract information from existing company databases and transform it to provide richer and more detailed knowledge. Second, a spatial way to estimate company networks based around their geographical location and industry type. Third, improved methods of estimating networks and applying methods to data provided by our collaborators. Last, a number of analytical techniques that can be applied to the transaction data we have to develop insight. A report and demonstrator have been produced for our collaborating partner, covering a number of use cases.
Exploitation Route Discussion with our partner re taking the work forward commercially did not result in concrete outcomes. Publications are being fed back to the data providing partner to garner take-up of methods.
Sectors Digital/Communication/Information Technologies (including Software),Financial Services, and Management Consultancy,Retail,Transport

 
Description SCRIBE Lloyds 
Organisation Lloyds Bank
Country United Kingdom 
Sector Private 
PI Contribution The development of novel business models, new product/service opportunities etc., based on an understanding of (transactional) client networks.
Collaborator Contribution Provision of sample banking transaction data.
Impact N/a
Start Year 2016
 
Description Accounting for network effects in credit risk modelling 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Other audiences
Results and Impact Talk to approximately 50 Doctoral Research Students and academics at STOR-i - a pioneering four-year doctoral training programme exploring ground-breaking approaches to Statistics and Operational Research, held at Lancaster University. The intended purposes of the talk were to: (a) Explore how transaction data can be used for credit risk modelling and the advantages that this may bring in terms of predictive power;(b) present a correlated mixed probit model for capturing dependencies that are either not observed or not accounted for by the transaction network; and (c) discuss the development of an efficient expected-maximisation algorithm for penalised inference. The performance of the methods were demonstrated on a large sample of accounts for small and mediumsized enterprises in the UK.
Year(s) Of Engagement Activity 2020
URL https://www.lancaster.ac.uk/stor-i/
 
Description Accounting for network effects in credit risk modelling 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact The activity took place at the European Commission and its intended purpose was to disseminate work on novel credit risk modelling to members of the Commission. The audience were provided with: (a) An overview of the importance of interfirm financial links in determining a company's performance (few studies have incorporated proxies for interfirm links in credit risk models, and none of these use real financial transactions); (b) an overview of a novel credit risk models for small and mediumsiz
Year(s) Of Engagement Activity 2019
 
Description SCRIBE Credit Risk Ontology Workshop (CROW) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Industry/Business
Results and Impact The CROW workshop sought shed light on the ontological nature of credit risk by bringing together domain experts from the financial industry with researchers in formal ontology. Approximately 25 people attended the workshop from financial organisations such as Lloyds Banking, Credit Suisse, Equifax. The workshop sparked interest and questions in a network-based understanding of credit risk and the use of ontology as a way of enhancing data semantics in risk models. Concrete outcomes come in the form of full-project participation by Lloyds and the development of an additional relationship with Equifax - Non-Disclosure Agreements have been signed with both.
Year(s) Of Engagement Activity 2015
URL http://www.scribe.org.uk/event/credit-risk-ontology-workshop-crow/
 
Description SCRIBE Networks and Estimation of Non-Linear Models with Spatial Dependence 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact The aim of the event was to discuss network-based approaches to identify, measure, and model interdependence among economic agents. The primary audience was doctoral students, researchers and academics alike - an international keynote was provided on the estimation of nonlinear models with spatial dependence. There was a good degree of discussion during the event and, as a concrete outcome, a working relationship (on the SCRIBE project) was developed with Professor Daniel McMillen of the University of Illinois.
Year(s) Of Engagement Activity 2015
URL http://www.scribe.org.uk/event/workshop-on-big-data-networks-and-the-estimation-of-nonlinear-models-...
 
Description SCRIBE Round Table on Big Data and Statistical Applications 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact A round table discussion between academics and industry participants, which sparked debate around big data and (risk) modelling.
Year(s) Of Engagement Activity 2017