CR-ONTO: Ontology-Based Reasoning for Credit Default Classification
Resumo
Context: Credit defaults are operations less explored by financial institutions compared to performing and non-performing loans. This is because recovering credit default is often ineffective and involves high collection costs. Problem: The literature presents studies that describe ontologies to represent knowledge about loans, but they do not address the conceptualization of default. Furthermore, the ontologies do not explore the possibility of recovering these credits in default condition. Solution: This article presents CR-Onto, an ontology designed for the classification of loans in general, with an emphasis on default. CR-Onto can represent the knowledge of the context involving the credit default and respond to questions suggesting whether a credit default can be settled. SI Theory: CR-Onto is the result of an interdisciplinary study that integrates concepts from finance, computing, and knowledge representation. By organizing data into an ontology, performing inferences, and executing queries, the data is transformed into information that can help credit advisors at a financial institution make better decisions. Method: The ontology has axioms and semantic rules used to provide queries and inferences about its instantiated base. The results were analyzed using a quantitative approach. Summary of results: A synthetic dataset based on a financial institution was constructed to simulate the lifecycle of 5 banking clients. This dataset populated 35 distinct instances of the CR-Onto ontology. This enabled the execution of DL Query to answer 6 competency questions. Contributions and Impact in the IS area: CR-Onto is capable of identifying which credit operations have a higher probability of being resolved. A specialist can utilize these classification results as part of an Information System within a decision support platform.
Referências
Robert Ahn, Sam Supakkul, Liping Zhao, Kirthy Kolluri, Tom Hill e Lawrence Chung. 2021. A goal-oriented approach for preparing a machine-learning dataset to support business problem validation. Em 2021 IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology Congress. IEEE, 282–289. DOI: 10.1109/DASC-PICom-CBDCom-CyberSciTech52372.2021.00057.
BACEN. 2013. Central bank of brazil. [link]. Accessed in may 2022. (2013).
Rodrigo Bavaresco, Yutian Ren, Jorge Barbosa e G.P. Li. 2024. An ontology-based framework for worker’s health reasoning enabled by machine learning. Comput. Ind. Eng., 193, C, (jul. de 2024), 16 pages. DOI: 10.1016/j.cie.2024.110310.
Asghar Beytollahi e Hadis Zeinali. 2020. Comparing prediction power of artificial neural networks compound models in predicting credit default swap prices through black–scholes–merton model. Iranian Journal of Management Studies, 13, 69–93. DOI: 10.22059/IJMS.2019.276260.673534.
Brazil. 2002. Law n. 10.406, january of 2002. art 206, incised 5. official diary of the federative republic of brazil. Accessed in march 2022. (2002). [link].
Carol Carson e Stefan Ingves. 2003. International monetary fund - financial soundness indicators - background paper. Accessed in may 2022. (2003). [link].
EDM Council. 2020. The financial industry business ontology. Accessed in july 2023. (2020). [link].
Eduardo M. da Silva e Jorge L. V. Barbosa. 2023. Machine learning and credit default: a systematic literature review and taxonomy. International Journal of Business Information Systems, (mar. de 2023), 22 pages. DOI: 10.1504/IJBIS.2023.10055817.
Yosi Lizar Eddy e Engku Muhammad Nazri Engku Abu Bakar. 2017. Credit scoring models: techniques and issues. Journal of advanced research in business and management studies, 7, 2, 29–41. [link].
Gaia Gambarelli, Aldo Gangemi e Rocco Tripodi. 2023. Is your model sensitive? spedac: a new resource for the automatic classification of sensitive personal data. IEEE Access, 11, 10864–10880. DOI: 10.1109/ACCESS.2023.3240089.
Michael Gruninger. 1995. Methodology for the design and evaluation of ontologies. Em Proc. International Joint Conference on Artificial Intelligence’95, Workshop on Basic Ontological Issues in Knowledge Sharing. [link].
Gilson Augusto Helfer, Adilson Ben da Costa e Jorge Luis Victória Barbosa. 2025. Soilbr-onto: an ontology for soil fertility management and classification in brazil. Applied Ontology. DOI: 10.1177/15705838251315571.
Yiping Huang, Longmei Zhang, Zhenhua Li, Han Qiu, Tao Sun e Xue Wang. 2020. Fintech credit risk assessment for smes: evidence from china. IMFWorking Papers, 20, (set. de 2020). DOI: 10.5089/9781513557618.001.
Evangelos Kalapodas e Mary E Thomson. 2006. Credit risk assessment: a challenge for financial institutions. IMA Journal of Management Mathematics, 17, 1, 25–46. DOI: 10.1093/imaman/dpi026.
Johannes Kriebel e Lennart Stitz. 2022. Credit default prediction from user-generated text in peer-to-peer lending using deep learning. European Journal of Operational Research, 302, 1, 309–323. DOI: 10.1016/j.ejor.2021.12.024.
Natasha Noy. 2001. Ontology development 101: a guide to creating your first ontology. [link].
Ashish Singh Patel, Giovanni Merlino, Antonio Puliafito, Ranjana Vyas, OP Vyas, Muneendra Ojha e Vivek Tiwari. 2023. An nlp-guided ontology development and refinement approach to represent and query visual information. Expert Systems with Applications, 213, 118998. DOI: 10.1016/j.eswa.2022.118998.
Baudino Patrizia, Orlandi Jacopo e Zamil Raihan. 2018. Bank for international settlements - financial stability institute. the identification and measurement of non-performing assets: a cross-country comparison. Accessed in April 2022. (2018). [link].
Kai Petersen, Robert Feldt, Shahid Mujtaba e Michael Mattsson. 2008. Systematic mapping studies in software engineering. Em 12th international conference on evaluation and assessment in software engineering (EASE). BCS Learning & Development. DOI: 10.14236/ewic/EASE2008.8.
Lucas Pfeiffer Salomão Dias, Henrique Damasceno Vianna, Wesllei Heckler e Jorge Luis Victória Barbosa. 2024. Identifying chronic disease risk behaviors: an ontology-based approach. iSys - Brazilian Journal of Information Systems, 17, 1, (jun. de 2024), 7:1–7:31. DOI: 10.5753/isys.2024.3762.
Reuters. 2020. Bank of brazil sells $545.4 million in loans to btg pactual. Accessed in may 2022. (2020). [link].
Open Risk. 2021. Non-performing loan ontology. Accessed in june 2023. (2021). [link].
Yuriy Ivanovich Sigidov, Marina Aleksandrovna Korovina, Aleksander Ivanovich Trubilin, Viktor Vilenovich Govdya e Nadezhda Konstantinovna Vasilieva. 2016. Creation of provision for doubtful debts. International Journal of Economics and Financial Issues, 6, 4, 1542–1549. [link].
Sanju Kumar Singh, Basuki Basuki e Rahmat Setiawan. 2021. The effect of non-performing loan on profitability: empirical evidence from nepalese commercial banks. The Journal of Asian Finance, Economics and Business, 8, 4, 709–716. DOI: 10.13106/JAFEB.2021.VOL8.NO4.0709.
Kristian Stancin, Patrizia Poscic e Danijela Jaksic. 2020. Ontologies in education–state of the art. Education and Information Technologies, 25, 6, 5301–5320. DOI: 10.1007/s10639-020-10226-z.
Jin Xiao, Yadong Wang, Jing Chen, Ling Xie e Jing Huang. 2021. Impact of resampling methods and classification models on the imbalanced credit scoring problems. Information Sciences, 569, 508–526. DOI: 10.1016/j.ins.2021.05.029.
Natalia Yerashenia e Alexander Bolotov. 2019. Creating an intelligent system for bankruptcy detection: semantic data analysis integrating graph database and financial ontology. Em Automated Reasoning Workshop 2019: Bridging the Gap between Theory and Practice. Middlesex University. DOI: 10.1109/CBI.2019.00017.
Samreen Zehra, Syed Farhan Mohsin Mohsin, Shaukat Wasi, Syed Imran Jami, Muhammad Shoaib Siddiqui e Muhammad Khaliq-Ur-Rahman Raazi Syed. 2021. Financial knowledge graph based financial report query system. IEEE Access, 9, 69766–69782. DOI: 10.1109/ACCESS.2021.3077916.
