improve english and be in business style. "Tks for the walk through yesterday. I have a couple of questions –
(1) SCB has a rule-based MSA gamer model. Any considerations on the swap analysis on these 2 models?
(2) The model is trained and validated in a relatively interest stable environment. Are there some tests in changing/low rate environment?
(3) The docs mentioned some features’ PSI are high. Which ones? Are those balance or amount related?
(4) Model accuracy is ~50%. Any business justifications?
(5) Given TD elasticity model exists, any considerations of the overlay of 2 models?
"
Thank you for the walk-through yesterday. I have a few questions:
Just wanna let you know that OW asked me to review the model, from the model governance and approval perspective. Plz see my detailed feedbacks in the forwarded email below; here is the summary.
• Model quality performance: From the statistical standpoint, the model has passed the Group’s quality metrics (e.g. GINI, PSI…); although a few features’ PSI is high, which should be closely monitored. However, the model accuracy is only ~50%. For your attention on the rest ~50% mis-clasifications and risk.
• External/changing environment: the model was trained and validated in a relatively stable rate environment. It would be beneficial to observe/validate in a changing/low rate environment.
• Overlap or swap analysis with other models, like OW’s TD pricing model; our own MSA-gamer model, I believe it would be beneficial to know the analysis so to set the pricing in various use cases.
I am aware of a walk through meeting with you on this Thursday. Happy to discuss there; feel free to let me know if you have any questions in advance.
I would like to inform you that OW has requested me to review the model from the perspective of model governance and approval. Please refer to the detailed feedback provided in the forwarded email below. Here is a summary of the key points:
Model Quality Performance: The model has met the Group's quality metrics from a statistical standpoint, such as GINI and PSI. However, there are some features with high PSI values that require close monitoring. Despite this, the model's accuracy is only around 50%, highlighting the need to address the remaining misclassifications and associated risks.
External/Changing Environment: The model was developed and validated in a stable interest rate environment. It is recommended to test and validate the model in a changing or low-rate environment to assess its performance under different conditions.
Overlap or Swap Analysis with Other Models: Considering the potential overlap or swap analysis with other models, such as OW's TD pricing model and our own MSA-gamer model, it would be beneficial to conduct an analysis to inform pricing strategies across various use cases.
I am aware of our scheduled walkthrough meeting this Thursday and look forward to discussing these points further. Please feel free to reach out if you have any questions beforehand.