引導式測驗 · 中等 等級
違約機率模型測驗
介紹 PD 模型目標、範圍、校準和驗證,同時將風險估計與借款人層面的確定性分開。
問題 1 的 8
What does a probability-of-default model estimate?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- What does a probability-of-default model estimate?
- A borrower’s guaranteed loss amount
- The current value of collateral
- Whether a borrower committed fraud
- The chance of a defined default event over a stated horizon
- Why must a PD model state its default definition and horizon?
- A model cannot use numeric features
- The estimate’s meaning depends on those choices
- The horizon determines a borrower’s identity
- It guarantees regulatory approval
- In the Basel IRB framework described in the guide, what is the reference PD horizon for specified corporate exposures?
- One day
- Five years in every case
- 一年
- The full life of the bank
- Which situation constitutes data leakage in a PD modeling workflow?
- A feature contains information that was unavailable at prediction time
- A borrower defaults before loan approval
- A model reports a probability
- A validation sample is too small
- Why can a random train-test split be misleading for default data?
- It always improves calibration
- It removes the need to define default
- It may hide time drift or leak information across periods
- It guarantees an unbiased sample
- What does calibration check?
- Whether a model uses a neural network
- Whether every borrower has the same score
- Whether the institution has collateral
- Whether predicted rates align with observed default frequencies
- Does a high PD score prove that an individual borrower will default?
- No; it is an estimate within a defined setup
- Yes, the score is a certainty
- Only if the model is machine learning
- Yes, under all Basel approaches
- Which statement about Basel IRB requirements is accurate?
- 它們適用於所有建立任何信用評分的機構
- 它們是針對使用內部評級法(IRB)的銀行的特定監管要求
- 它們只是模型供應商的建議
- 它們取代了消費者保護法