Prediction of Factors Influencing Social Performance of Indian MFIs using Machine Learning Approach

Abstract

The aim of the current work is to predict the impact of MFI specific internal factors on the social performance of Indian microfinance institutions (MFIs) by using machine learning techniques. Social performance index (SPI) is designed by taking data of 73 Indian MFIs for 10 years with the help of an indexing technique where six different factors (operational self sufficiency, number of women borrowers, number of rural borrowers, gross loan portfolio, average loan balance per borrower / GNI per capita and cost per borrower) representing different dimensions of functioning of MFIs are considered. The data is taken from MIX data repository. Pooled OLS regression model is used for analyzing impact of various MFI specific factors on SPI. For predicting the SPI, Artificial Neural Networks (ANN) machine learning model is considered that takes all independent variables as input. The results of regression model indicate that size, legal status, outreach and service provisions significantly affect SPI. ANN analysis result indicates that social performance of MFIs gets determined by MFI specific internal factors. The experimental result indicates that the proposed ANN prediction model is providing better result for predicting the SPI. The findings suggest that MFIs can contribute for development of the society by adopting suitable policies keeping in view certain internal factors.