AI driven credit risk assessment based on climate change factors

Our data science team works on integrating machine learning capabilities into the credit lifecycle and climate risk management system to substantially improve risk management for financial service providers.

About the Client

YAPU Solutions is an innovative FinTech startup for farmers focusing on the populations most vulnerable to the climate and biodiversity crisis in developing countries. Due to various climate-changing events caused by human activity impacting weaker rural markets, YAPU recognizes the urgent need for resilient financial solutions that would help small land-holding farmers to streamline production by increasing working capital. With Unicsoft support, the company provides a digital platform for loan management and an operating system for resilience focused climate finance and increased impact reporting for financial institutions worldwide – as recognized and supported by multilateral development banks such as IDB Lab and UN IFAD or large commercial banks such as BNP Paribas.

Unicsoft has led the technological development of a YAPU designed credit lifecycle management system. Based on several years of triumphant product introduction, the client is looking to complement its data gathering capacities with AI-powered modules that further enhance risk management and improve operational efficiency for financial institutions.

Solution Highlights

YAPU’s cutting-edge system is increasing loan origination efficiency and capable of lowering credit default for financial institutions due to an increased transparency while delivering a better view on climate- and nature-related risks in credit assessment. The solution provides data insights on credit risks and cash flow projections, based on analyzing a spectrum of external factors, including climate threat exposure and sensitivity, production or asset characteristics (e.g. for housing, crops and animals) or price tendencies.

Business Goals

YAPU is powering up its existing system and analyzes the impact of applying ML to predict the probability of a borrower defaulting on his/her loan. Loans to vulnerable borrowers in developing countries are different to traditional business loans in highly formalized financial sectors – they are characterized by high levels of informality and market volatility. There are a number of different qualitative and quantitative risk factors to measure before financing higher risk loans for these populations in highly informal environments.

Project Scope. Step 1.

Unicsoft has charted a comprehensive roadmap, outlining the integration of AI functionality into the existing credit management solution from the ground up. With YAPU, we have conducted a thorough feasibility study, co-financed by the German government, defining four potential data use cases, identifying what data categories need to be gathered and processed, and designing the conceptual outlines on algorithms for future model training.

Credit risk assessment

Before, the system relied on expert-driven reference data on external factors and a complex set of formulas to calculate potential risks. So far, actual payment performance data could not be integrated in a standardized way. The new ML model is tasked with assessing credit risks using historical payment performance data and internal data sets on the business reality and external factors alike, including numerous climate factors.

Expected results

Financial institutions will receive more detailed credit risk assessment reports based on several data sources and comprising credit ratings, sector and activity risks, cash flow projections as well as physical climate risk estimations. As a result, financial service providers will be able to make more objective and data-driven decisions, disclosing physical climate-related financial risks and accelerating credit disbursement processes, while lowering credit defaults.

Enhancing the inclusion of standardized external factors

In YAPU’s current digital lending approach, front office personnel (e.g. loan analysts or client advisors) must manually select reference data sets for each new loan evaluation. These data sets are being gathered by specific initiatives, a lengthy and cost intensive exercise. Further, this manual and error prone process is far less accurate than ML-powered matching algorithms, and lacks automated approaches to review information relevance and quality.

Expected results

The proposed ML module will analyze data gathered by field personnel to automatically build the reference data sets. Further, the system will be able to automatically match reference data sets to individual loan evaluations, analyze mistakes and data accuracy, as well as to give recommendations for improvement.

The prioritization of adequate adaptation solutions

In order to address identified climate risks, specific adaptation solutions can be implemented by end borrowers and financed by YAPU client institutions. To identify the best climate adaptation options for an individual client and even express adequate recommendations is equally a cost and time intensive process, prone to human error. Replacing this expert-driven process by an ML module ensures the credit process automatically includes verification of investment, together with productivity and climate resilience benefits.

Expected results

The new ML model will analyze all the major climatic factors affecting production and payment performance, and come up with proposed adaptation solutions to mitigate those identified risks. When preparing a loan evaluation, the system will advise securing additional lines of credit for meeting the climate needs that may arise in the future, thus helping to improve credit scores. 

Enhanced operational efficiency

In the above described informal MSME finance settings in developing countries, financial institutions’ front office personnel is usually constantly visiting clients at their private or business premises. Transport between and prioritization of visits is hence of utmost importance and in traditional settings managed by supervisory functions within the institutions – manually.

Expected results

Machine learning can optimize the sequence of subsequent visits for different purposes, like loan evaluation, loan work-out of non performing loans or business development activities. Linking the outcome of specific activities to their ML-driven prioritization can enhance field personnel efficiency and effectiveness.

Project Scope. Step 2.

Currently, Unicsoft works on covering the use cases 2 and 3 into a merged uses case of climate risk assessment and adaptation solutions recommendation system. First prototypes have been developed, while YAPU is aligning its data gathering routines and processes for existing and new clients across the entire platform. This development was and is following the established ML model development process:

  • defining business requirements
  • analyzing available data
  • collecting & preparing data
  • feature modelling
  • model’s performance evaluation & optimization.

Expected Outcomes

Future ML modules will be seamlessly integrated into the existing YAPU system and resolve the challenges mentioned above. These modules are projected to enhance the credit risk assessment system, optimize credit scoring & management processes, and make the system more efficient to use.
Elevate your credit risk assessment with AI: experience unparalleled accuracy & risk control
Contact the expert