2021 | Finance/ Credit Companies | Nigeria | Promising | SDG1 | SDG10 | SDG3
Novel Machine Learning Approach for Credit Risk assessment using Non-Traditional Data Sets

1. General


SDG 1: No Poverty

SDG 3: Good Health and Well-being

SDG 10: Reduced Inequality


Finance / Credit Companies

2. Project Details

Company or Institution

Vittas Inc


Novel Machine Learning Approach for Credit Risk assessment using Non-Traditional Data Sets

General description of the AI solution

Vittas is part of the IBM HyperProtect Cohort – and was recommended to apply through the program. Vittas intends to use Machine Learning to execute micro-segmentation based on customer behaviors rather than solely on credit history identifiers. Machine Learning will “train” models based on behavioral and traditional data sources to enhance the predictive power of the credit models and has shown consistent accuracy and capacity to capture non-linear relationships characteristic of credit risk. This research will allow Vittas to innovate and introduce new products and services within emerging markets (initially Nigeria)– where there is a lack of a centralized infrastructure that monitors consumer credit history, which has led to risk averse lending practices minimizing SMB’s access to affordable working capital. The credit to GDP ratio in Nigeria is 11% compared to 191% in USA. By providing access to affordable loans – SMB’s (the backbone of an economy) will be able to grow their businesses and hire more people from their local communities. This will increase quality of life from the ground up. Initially Vittas has focused on the healthcare market – ensuring hospitals/pharmacies have money to purchase medications to reduce inventory stockouts which occur 30-50% of the time in Nigeria. Vittas is the first stage of the development process.




Vittas Inc

3. Aspects

Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.

Vittas’ business bears with a lot of challenges in data scarcity and data integrity. Depending on the levels of data scarcity in Vittas’ business, several advanced artificial intelligence methodologies such as transfer learning, meta learning and unsupervised domain adaption, which have the flexibility of incorporating non-traditional data sets from other domains, are adopted to address the small data issue Vittas is facing in emerging markets. To deal with the data integrity in Vittas’ current business setting, distributionally robust optimization techniques are developed to hedge against the risk of data corruption. Active learning and online learning techniques will also be employed to facilitate Vittas’ ongoing data collection effort. Finally, novel loan-pricing optimization models is being developed leveraging the powerful artificial intelligence methodologies.

Some of the artificial intelligence methodologies have found successfully applications such as natural language processing, gaming, autonomous vehicles, digital agriculture. Yet, applying these artificial intelligence methodologies to address data scarcity and integrity is new in the kind of business Vittas is facing. Integrating these artificial intelligence methodologies into Vittas’ loan-pricing optimization models along the the online-learning-framework developed in Ye et al. (2020), a research paper by my research group, is very novel and has seldom been addressed even in the academic literature. Vittas has outlined a clear and detailed blueprint for the integration and its implementation and demonstrated quite promising initial results.

IP is being protected using trade secret to ensure Vittas does not need to disclose the details of the model to anyone. Vittas is taking the necessary steps to ensure the IP is protected and kept secret.

Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.

Vittas has disbursed $300,000 USD in loans impacting over 1000 individuals and 20 companies. During the pandemic, Vittas customers have grown their businesses and provide life saving medications due to Vittas (testimonials are available).

Vittas will use the online-learning-framework developed in Ye et al. (2020) and deployed in a large scale online platform to test accuracy. NPS scores, SMB-revenue growth, employment numbers will be used to assess effectiveness alongside internal default-rates.The loan-pricing-problem is related to the work of Ye et al. (2020) in four aspects: 1) The predicting parts are both built on advanced machine learning models; 2) Exploitation: in addition to the prediction, Vittas must solve non-trivial optimization problems; 3) Exploration: data scarcity calls for collecting more diversified-data, which will be fed into ML models for training, (the labeled default-loan-data for different rates in this setting, and the labeled click-data for different new-ads) 4) Both problems require jointly/repeatedly balance the exploration & exploitation.
Though online-learning approach has not been used in the business setting, set to be addressed in this proposal, it is encouraging to see that it has been successfully implemented in an online-advertising-platform by Ye et al. (2020). The assumption is similar ideas would be powerful for the loan-pricing-optimization-problem Vittas is facing with the credit-invisible market segment. This would democratize access to working capital.

People need access to affordable capital to increase earning potential. This problem is severe in emerging markets where such credit-scoring infrastructure doesn’t exist. Fixing this will reduce the unemployment rate by increasing money flow to unserved market-economic-segments. The loans will allow businesses to grow which will trigger economic growth and an enhanced standard-of-living at the local and national levels
By creating a ML model that works in this credit environment – the use of AI will exponentially increase to help solve global-financial-inclusion issues.

Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.

Evidence for impact is testimonials from pilot customers and impacts on their businesses. Even during the pandemic these businesses were able to expand and grow – due to Vittas’ ML model allowing for risk assessment in a data scarce environment. The resulting default rate was 0% – showing that the model was also effective; though a larger sample size is needed for validation. The ML model will improve scalability and access by Vittas leasing the model to financial institutions to use, similar to FICO, for emerging markets. By using the ML model that is created, the model will be continually iterated and improved to better fit the markets, verticals and regions it is in. Because the lack of lending is a 5 trillion dollar global problem – the emergence of our technology will allow for large scale impact across the globe with stakeholders in all sectors. The initial target market is Nigeria – Vittas is developing/incorporating transfer learning to ensure global adoption. This will support he emergence of AI for financial inclusion across the globe. Vittas has 20 early customers – our product currently impacts 800 individuals. This number will expand after Vittas begins to scale. The initial target community is Nigeria – with 42 million SMBs and a population of 201 million. Vittas is GDPR compliant.

Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.

Vittas applications are based on social impact and factor/value ethical and equitable application. Vittas is using ML to provide lowest in market interest rates for non-collateralized loans. Trustworthiness of our ML solution is critical due to the nature of our product. For adoption to occur Vittas’ ML solution must: 1) be lawful 2) be robust (data to back how accurate model is). Lastly, Vittas has found that by incorporating ethical principles – Vittas is able to create a sustainable lending ecosystem that creates a recurring customer base. It makes more money to be fair because SMBs grow and increase their business with Vittas (Vittas has seen this take place with our Pilot customers). The goal of Vittas tech is to provide financial inclusion to SMBs especially those historically excluded from financial services (women, people of color, base of pyramid). Our mission statement is to provide financial inclusion for the underserved – and all of our decisions are taken to help achieve this goal.


International Research Centre
on Artificial Intelligence (IRCAI)
under the auspices of UNESCO 

Jožef Stefan Institute
Jamova cesta 39
SI-1000 Ljubljana



The designations employed and the presentation of material throughout this website do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country, territory, city or area of its authorities, or concerning the delimitation of its frontiers or boundaries.

Design by Ana Fabjan