Measuring behavioural change has traditionally been a challenge, yet there is a way to overcome

Continuous learning and improvement are core principals of agile working. Successful agile transformations have shown the value of monitoring process, evaluating behavioral change and its impact on performance.

For lending, risk models are required to have an agility in order to manage loan portfolios well especially times like now under the economic slump. Deterioration in credit qualities has become more evident in Hong Kong’s credit market.

The first step is to measure behavioral change and model accuracy consistently and continuously. In fact, measuring behavioral change has traditionally been a challenge, but there is a way to overcome this challenge. Let’s take it further.

Risk monitoring on credit scoring model is a must-have function for lenders to make sure your risk model is working, otherwise your lending cannot react speedily to the changes in environment or risk indicators. In order to implement risk monitoring, it is inevitable to deploy the tool. This is where CREDI AI can help.

In lending, we don’t have that much time to conclude a risk model is working. Consumer lending portfolios can normally take up to 18-24 months for lenders to acknowledge that the original prediction was accurate, while changing is coming constantly from a rapid deceleration in credit quality due to a pandemic. To ensure safe and efficient operation of machine learning models, CREDI AI also deploys monitoring features along with predictive models.

The short video helps you to understand what our monitoring features are about. You can find main functions from the list below as well.

• Score Distribution
  These charts are auto tracking features to grasp score distribution of borrowers. Users can identify credit risk trends of borrowers and easily grasp a tendency from the time-series data.

• Risk Analysis
  The following two risk analyses enable to analyze default risks and model accuracy by comparing model data and actual performances.
  1. Default risk analysis per credit grade
  2. Data table (which shows comparison data)

• Value Chart
  Various charts show distributions of variables. Users can easily figure out differences of data distributions between models and actual results. These charts allow them to detect variable changes such as a change of borrower attributes.

• Value Analysis
  Outliers will be listed up in the table once the software detects them. The occurrence frequency of outliers is a key performance metrics to discover a behavioral change.

Monitoring these features are continuously made available to our lender customers so that they can find and review risk indicators based on data. At CREDI AI, we set up regular review sessions with customers to look over model performance and what monitoring data are telling us.

While every lending business has been affected by Covid-19, some lenders are still using a risk model trained by the pre-pandemic data so that their model accuracy has been deteriorated. The monitoring live data can notify us if there’s an impact on the model accuracy. If this is significant enough, it may be time now to rebuild that model to improve accuracy.

As monitoring has become more important in the times like now, agility and resilience of re-building the model are required for lenders especially if your risk model was trained only by the pre-pandemic data.

We are happy to discuss about how the risk monitoring enables lenders to keep monitoring on model accuracy. If you are interested in the monitoring features, send us an email to: enquiry@crediai.com !