Facilitating Credit Access for Unbanked Populations in Africa through Machine Learning

Challenge Completed!

Despite data availability limitations, the team pre-processed various financial data sets to develop two machine learning models with more than 95 percent accurary. The models predict the default “Loss Given” a lender may incur and a minimum and maximum loan amount the lender may consider.

As part of Omdena´s AI Incubator, the partner for this project is the Seedstars awarded startup Toju Africa, which is on the mission to use technologyto facilitate access to financial services for underserved populations in Africa.

The problem

Many populations in Africa are still struggling to get access to financial services. Toju Africa aims to provide affordable monetary assistance to the last mile through Thrift collectors, savings clubs, and local co-operatives. Toju helps local financial service providers do more by keeping better records and adding income streams. Accessing credit for the underserved population who utilize these local financial service players can be a challenge because their data is manually stored and can’t be accessed by formal loan companies.

Toju has digitized the data collection and storage for the local financial service players and is partnering with lenders to use these data to inform loan decisions.

Their mission is to allow underserved populations to access financial services, especially loans, and to build a predictive model to suggest loan capacity using transactional data from informal financial service providers.

Snapshot: Toju Africa dashboard

The project results

Specifically, the goal was to apply machine learning to predict loan capacity in contexts with no or little access to standard credit scoring facilities such as credit bureaus. The team produced two machine learning models with an accuracy score of more than 95 percent with the following two critical predictions: