ABSTRACT
A database of financial statements is a financial infrastructure for conventional banking systems to assess the creditworthiness of micro-, small-, and medium-sized enterprises (MSMEs). A crucial assumption is that MSMEs prepare accurate financial statements. In addition, financial statements are generally updated only once a year, which means that the data is not timely enough to evaluate the creditworthiness of MSMEs. Furthermore, any MSME engaging in business activities is likely to have a deposit account, making it possible to collect data and build a model for MSMEs based on deposit data. This study introduces the application of machine learning to the credit-scoring model to achieve a high-performing credit-scoring tool based on a large amount of complex deposit data, while complementarily evaluating traditional financial statements. Furthermore, there exists room to improve the model’s accuracy by revising the model building by applying new AI technology and alternative data.
