The mortgage servicing industry has long loan portfolio lifecycles. During this time borrowers may face a variety of life circumstances that can prevent them from keeping up with mortgage payments. Owing to this, about 2-4% of loans tend to go into default. Infosys’ Mortgage Default Prediction System leverages Artificial Intelligence and Machine Learning (AI/ML) to identify potential loan defaulters and take preemptive action.

Our solution can help mortgage companies flip the industry trend in handling loan defaults by moving from reactive to proactive mode. With a high degree of accuracy in predicting loans that could go into default, clients can rely on our solution to take corrective action with considerable year-on-year savings.

Bank – Mortgage Loan Data

  • Loan Portfolio
  • Customer Information

Public Data Bureau of Labor Statistics

  • Unemployment Rate at Job Location
  • Employment rate at Job Sector

Mortgage Investor Data

  • Fannie Mae Quarterly Investor Data
  • Freddie Mac Quarterly Investor Data
  • Leading Credit Bureau (Potential partnership)

Probability to Default

  • Score between 1 to 100 on every loan
  • Higher score = higher chance of default

Factors affecting Default

  • Due of Passed Missed payments
  • Due to LTV ratio
  • Due to employment sector not doing well


  • Manual remediation – offer mods
  • Potential automation – based on manual remediation patterns

Enabling Default Prediction Using Artificial Intelligence and Machine Learning

Talk to our experts

Infosys’ Mortgage Default Prediction System uses a combination of available industry data, artificial intelligence and machine learning algorithms to provide the mortgage industry with a fairly accurate view of potential defaulters.

This is how it works. The lending bank has the customer’s loan portfolio. This includes information about any loan payment defaults already recorded in the system. In addition, our solution uses public APIs to collate historic information from mortgage investors such as Fannie Mae and Freddy Mac. Lastly, macro-economic influencers such as unemployment statistics in the area where a borrower lives are gathered from government sources.

This intelligence is passed through the machine learning model to generate a default score for each borrower. On a scale of 1 to 100, the higher the score, the greater the chances of the borrower defaulting on the loan payment.

Clients can use this score to negotiate remediations with the borrowers. This data can in turn be used to automate remediations over time.

Infosys’ Mortgage Default Prediction System: A future-proof, automated solution for mortgage default prediction

Up until now, there has been no single source of truth that could aggregate borrower and industry data to accurately predict defaults.

Our solution combines modern, scalable technology to provide a future-proof solution that uses:

  • Deep neural networks with multiple layers to train the machine learning model
  • Explainable AI to highlight potential reasons for default for better risk analysis
  • Public APIs from the US government’s Bureau of Labor Statistics to gather unemployment data at job sector as well as the metropolitan statistical area levels
  • Correlation analysis to analyze the huge amount of data from mortgage investors Fannie Mae and Freddie Mac

Challenges & Solutions

Without the ability to predict defaulters, mortgage servicing companies lose millions of dollars each year.

Infosys’ Mortgage Default Prediction System:

  • Provides high level of accuracy in predicting loan defaulters
  • Helps clients take proactive action and prevent defaults
  • Can save millions of dollars for large mortgage lenders

Although a lot of data is available in the public domain, historically no aggregated view of the data has been available.

Our solution uses public APIs to leverage:

  • Data from external sources such as the Bureau of Labor Statistics to fine-tune the default score
  • Government Sponsored Enterprise (GSE) investor data from Fannie Mae and Freddie Mac that contains mortgage loan servicing data from all US banks

Our solution integrates multiple sources of data and provides a rich and intuitive user interface for slicing and dicing the data for easy analysis.

Some of its key features include:

  • Web interface that provides access to data using any device, anywhere
  • Loan analysis screen with borrower details, default score, unpaid amount, and other relevant details
  • Filters to view potential defaulters within a range of scores and potential reasons for default to help make the right business decisions
  • Analytics graphs to view data based on various parameters such as likelihood and primary reasons for default