Real-time classification of SMEs credit and risk ratings and the impact of financial indicators and payment behaviour
Tuovinen, Teemu (2020)
Tuovinen, Teemu
2020
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2020060416845
https://urn.fi/URN:NBN:fi:amk-2020060416845
Tiivistelmä
This thesis studies the applicability and the benefits of real-time data in the forecasting of financial risk and credit ratings of Finnish SMEs and the impact of payment behaviour to ratings. The research aims to analyse if the data of an accounting firm could improve reliability and accuracy when evaluating the probability of defaulting and the financial risk of a business. In order to achieve the objective, the development of the process, methods, and algorithm are needed. The development of an automated credit rating process is time-consuming and requires resources as accuracy and reliability are fundamentals of classification. Credit and risk ratings explain the overall financial status and risk of the entity, which makes it complicated to determine the accurate indicators of the financial status of a business. This thesis explains the theory behind the phenomenon of credit and risk rating methods when analytics and accounting are combined.
Both quantitative and qualitative elements have a significant role in the rating process and methods. Qualitative variables measure the subjectivity of the business operations as quantitative variables measure the subjectivity of the financials. Hypotheses were set according to the results of the past studies related to the matter, and data analysis authenticated the operability with the multinomial Pearson correlation coefficient and supervised logistic regression classification model. The relation of the results was analysed against the ratings of rating agencies and past research in order to evaluate the operability of the classification model in a business case scenario.
The results confirmed that real-time data increases the quality and accuracy of financial information, but also that the data can be applied in the rating processes. The impact of the additional information implemented in the rating methods was partly beneficial as an increase in payments made late was seen as an explanatory variable when evaluating the risk. The result indicates that debit entries of clearing account in relation to all transactions do not have an impact on the riskiness or creditworthiness of the entity. The results differ from the past research and guidelines of credit rating agencies in the sense that not all the financial variables examined are significant when evaluating the risk and creditworthiness of an SME. The rating process and algorithms built according to the results are a basis that will be improved into a Neural Network rating model in the future.
Both quantitative and qualitative elements have a significant role in the rating process and methods. Qualitative variables measure the subjectivity of the business operations as quantitative variables measure the subjectivity of the financials. Hypotheses were set according to the results of the past studies related to the matter, and data analysis authenticated the operability with the multinomial Pearson correlation coefficient and supervised logistic regression classification model. The relation of the results was analysed against the ratings of rating agencies and past research in order to evaluate the operability of the classification model in a business case scenario.
The results confirmed that real-time data increases the quality and accuracy of financial information, but also that the data can be applied in the rating processes. The impact of the additional information implemented in the rating methods was partly beneficial as an increase in payments made late was seen as an explanatory variable when evaluating the risk. The result indicates that debit entries of clearing account in relation to all transactions do not have an impact on the riskiness or creditworthiness of the entity. The results differ from the past research and guidelines of credit rating agencies in the sense that not all the financial variables examined are significant when evaluating the risk and creditworthiness of an SME. The rating process and algorithms built according to the results are a basis that will be improved into a Neural Network rating model in the future.