Predictive Modeling of Asset Growth and Liquidity Risk in Umurenge SACCOs Using Machine Learning Techniques

This study examines the application of machine learning techniques to predict asset growth patterns and evaluate liquidity risk in Umurenge Savings and Credit Cooperatives (SACCOs) in Rwanda. Given the pivotal role of SACCOs in fostering financial in-clusion and local economic development, understanding the interaction between asset expansion and liquidity stability is essential for sustainable financial management. Using historical financial and operational data, multiple predictive models are developed and comparatively assessed to classify asset growth behaviors and associated liquidity risk levels. A range of supervised learning algorithms, including tree based models, ensemble approaches, and neural networks, is applied to key financial indicators such as deposits, loan portfolios, repayment performance, and liquidity reserves. Model performance is evaluated using standard metrics, including accuracy, precision, recall, F1-score, AUC, G-mean, balanced accuracy, and Matthews correlation coefficient (MCC). The results demonstrate that ensemble based models outperform conventional approaches, with the Extra Trees classifier achieving the highest performance (98.75 % accuracy, AUC = 0.9993, MCC = 0.9813). Confusion matrix analysis indicates high classifica-tion accuracy, with minor misclassifications concentrated in the controlled growth category due to its intermediate characteristics. Overall, the findings highlight the effectiveness of machine learning in capturing complex financial dynamics and supporting data driven early warning systems for enhanced decision making and sustainability in SACCOs. To the best of our knowledge, this study represents one of the earliest empirical contributions to predictive modeling of asset growth and liquidity risk in Umurenge SACCOs using machine learning techniques in Rwanda.