Default Risks Prediction
In Tanzania, most smallholder farmers rely on sporadic rainfall to water their fields and most fishers rely on kerosene lamps for light on their boats. With high-quality equipment, these farmers and fishers could increase their productivity by up to 300 percent. But high upfront costs and a lack of rural distributors keep these solutions out of reach for more than 40 million Tanzanians.
61% of Simusolar’s customers live on less than $3.10 per day. With Simusolar’s equipment, customers have the potential to double their incomes within a year. The company has served 2,000 farmers to date, with plans to reach 10,000 customers by 2021.
Currently their largest issues are related to default risks of their customers. In the past they have collected quite a lot of demographic data about their customers and have judged how likely they are to default with the naked eye. This has become quite tedious and they have asked us to asses the default risk using modelling techniques. For this we already have quite a big data dump consisting of various demographic data such as how far from a river a farmer lives, how many kids he has, age, other professions, how much land they have, etc. that can be used as independent variables. Furthermore we have a large amount of payment data that we can use as independent variables.
Aside from credit risks they would like to ensure the pumps will remain intact as long as possible. The pumps have a few IoT sensors on them with which they would like to perform preventative maintenance. This means that customers that use their pump is such a way that it will not last very long could be contacted and told how to improve the longevity of their pumps.