CareEdge | Risk Solutions

Objective

To analyse and determine risk levels involved in credits, finances and loans.
Assist analysts with predictive risk analysis.
Increase effectiveness of risk management by measuring the riskiness of borrowers.

Pain Points

1
Collecting historical financial & non-financial data from various internal & external sources.
2
Slicing and segmenting data across different views and perspectives for accurate analysis and decision making.
3
Use of efficient feature extraction techniques for dimension reduction.
4
Perform model validation / cross-validation to get better & accurate credit rating.

Solution Given

1
Machine Learning techniques to analyse and predict the borrower's credit risk rating for one of the biggest banks in Sri Lanka.

Solution Highlights

1
Effective data ingestion techniques to collect data from various internal & external sources.
2
Better data visualization tools for correlation analysis, feature extraction and data aggregation.
3
Used models from probabilistic classifiers and ensembled learning to create classification algorithms.

4
Used various metrics to interpret the predictive validity of a model (e.g., R-square, mean squared error, sensitivity, goodness of fit, ROC, loss function, confusion matrix), and validated the models by using methods such as e.g. Cross-validation and bootstrap.
5
Periodic review of the model and train the model based on the latest data.

Key Benefits

Benefits of the Case Study to help you understand our Product and reach of Services in a more convenient way

With the proposed solution, the client was able to calculate the credit rating of its borrowers with an accuracy of around 88%. It helped the client to improve the credit monitoring process by predicting the customer’s delinquency and accordingly categorise them into good account (“one who pay on time) and bad account (“one who default) with a quick turnaround time