Credit decision tool using mobile application data for microfinance in agriculture

• Published in: 2017 IEEE International Conference on Big Data (Big Data) pp. 4714 - 4721
• Main Authors: Simumba, Naomi, Okami, Suguru, Kohtake, Naohiko
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-12-2017
• Subjects: Agriculture; decision support systems; Logistics; microfinance; mobile application; Mobile communication; Mobile handsets; Productivity; Support vector machines
• DOI: 10.1109/BigData.2017.8258519

Smallholder agriculture is vital for food security and poverty reduction. Poor, unbanked smallholder farmers face challenges in obtaining loans for agriculture due to missing financial credit history. Although credit decisions tools with alternatives such as big data from mobile phones currently exist, these are mainly targeted towards consumer credit solutions. There is need for research focused on applying such data to credit decisions in agriculture, considering the unique challenges of the field. This research proposes a credit decision tool using alternative data from a mobile phone application used for reporting agricultural activities. 213 users submitted 41613 reports of farm activity by 11336 farmers over a 5-year period. Users were classified groups based on personal information, association with farmers, and reports. Three methods are applied for analysis: multiple logistic regression, a support vector machine with a linear kernel, and a support vector machine with a radial basis function kernel. Three sets of features, selected with different methods, are used. The results are evaluated by accuracy, and Area Under the Receiver operating curve (AUC). The support vector machine with a radial basis function kernel shows the best performance on the original data set with an Area Under the Receiver operating curve (AUC) value of 0.983. Feasibility of applying each method in agricultural microfinance is considered. Finally, further development of the models using Geographic Information System analysis of geospatial data which affects agricultural productivity, such as soil type and water resources, is suggested to incorporate the ability of the borrower to repay loans from farm activity.