Microfinance institutions currently confront numerous financing challenges, particularly within the non-bank sector where risks abound. Each year, there is a notable incidence of borrowers defaulting on their microfinance obligations, resulting in substantial financial setbacks for these companies. Given the burgeoning volume of electronic data and transactional activity in the banking realm, data mining emerges as a pivotal strategic domain. Leveraging data mining techniques, valuable patterns and insights can be gleaned from vast datasets, thereby furnishing actionable information to mitigate risks associated with nonbank loans. This study employs data mining as a tool to extract pertinent insights from the credit data of microfinance companies, facilitating the construction of a model aimed at assessing borrower eligibility and identifying potential default risks. The study employs the open-source machine learning platform WEKA. This study uses data mining to develop a predictive model for microfinance institutions to enhance decision-making in client financing. By employing cross-validation and percentage splits (80-20, 70-30, 60-40), cross-validation showed slightly higher accuracy. The model performed excellently, especially with preprocessed data, highlighting the importance of data cleaning. The J48 proved to be the most effective algorithm, demonstrating superior accuracy. The study emphasizes the potential of using historical data to assess client credit status during financing approvals, reducing loan defaults, and supporting the growth of non-banking institutions

Data Mining Technique, Classification, Credit Risk, Non-Banking Sector, Microfinance, Fraud detection

The paper utilizes data mining to develop a predictive model, focusing on the loan histories of existing borrowers. This model aims to aid in comparing potential loan applications by identifying characteristics indicative of a good or bad loan record, drawing from credit background and demographic profiles. Emphasis is placed on the importance of preprocessing or cleaning data to achieve higher accuracy rates. The results obtained using the J48 algorithm are particularly noteworthy, with a correctly classified instances rate of 99.7901%. Additionally, the preprocessing stage can reveal patterns useful for identifying target loan markets, devising income-enhancing strategies, reducing default risk, and improving loan products

Sameh Ali, Atef Raslan and Lamiaa Fattouh (2024). Model for the Prediction of Default Risk of Funding Requests Using Data Mining. International Journal on Emerging Technologies, 15(2): 05–12.