Prediction of SACCOS Failure in Tanzania using Machine Learning Models

Abstract

Savings and Credit Co-Operative Societies (SACCOS) are seen as viable opportunities to promote financial inclusion and overall socioeconomic development. Despite the positive outlook for socioeconomic progress, recent observations have highlighted instances of SACCOS failures. For example, the number of SACCOS decreased from 4,177 in 2018 to 3,714 in 2019, and the value of shares held by SACCOS members in Tanzania dropped from Tshs 57.06 billion to 53.63 billion in 2018. In particular, there is limited focus on predicting SACCOS failures in Tanzania using predictive models. In this study, data were collected using a questionnaire from 880 members of SACCOS, using a stratified random sampling technique. The collected data was analyzed using machine learning models, including Random Forest (RF), Logistic Regression (LR), K Nearest Neighbors (KNN), and Support Vector Machine (SVM). The results showed that RF was the most effective model to classify and predict failures, followed by LR and KNN, while the results of SVM were not satisfactory. The findings show that RF is the most suitable model to predict SACCOS failures in Tanzania, challenging the common use of regression models in microfinance institutions. Consequently, the RF model could be considered when formulating policies related to SACCOS performance evaluation.