Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ
AI trong bài kiểm tra dự báo ngân sách công
Kiểm tra dự báo ngân sách AI: mục tiêu và tầm nhìn, xác thực theo thời gian, giả định minh bạch, phạm vi không chắc chắn, đường cơ sở và trách nhiệm giải trình trước công chúng.
Câu hỏi 1 của 8
A revenue model is trained on five years of records. Which test best approximates forecasting next year?
Xem trước mọi câu hỏi8 câu hỏi
Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.
- A revenue model is trained on five years of records. Which test best approximates forecasting next year?
- Randomly split rows from all six years
- Evaluate only on the training records
- Train on earlier periods and evaluate on a later period kept out of model fitting
- Choose the year with the lowest error after viewing all results
- A model predicts annual tax receipts. Why must the office state its target and horizon?
- The target and time horizon define what the estimate means and how it can guide decisions
- They make an uncertain forecast certain
- They determine which political priorities should be funded
- They eliminate data revisions
- A tax-law change breaks a historical relationship in the data. What is the main risk?
- The model automatically learns the new law before it takes effect
- The change improves accuracy by adding more data
- The forecast horizon becomes irrelevant
- The model may extrapolate a pattern that no longer describes the revenue system
- Why compare a complex forecast with a simple baseline?
- To make the complex model appear more accurate
- To determine whether added complexity improves useful performance over a transparent reference
- Because a baseline guarantees the correct budget
- To avoid testing on future periods
- Which presentation best communicates uncertainty to budget decision-makers?
- Publish only the model’s most likely number
- Hide sensitivity analysis to prevent debate
- Show a central estimate with plausible ranges or scenarios and explain key assumptions
- Convert the estimate into a guarantee
- A random train-test split gives excellent results on monthly revenue data. What concern remains?
- The result proves the budget will be balanced
- Random splitting ensures a realistic future simulation
- The model has no need for a baseline
- Future observations may have leaked into training, making the test unlike a true forecast
- A city sees its average error fall but the model repeatedly misses during downturns. What should analysts report?
- Performance by relevant conditions, including downturns and large misses
- Only the overall average error
- Số tham số của mô hình
- That no forecast can be evaluated
- Who remains responsible for selecting spending priorities after an AI forecast is delivered?
- The forecasting model because it produced the estimate
- Public officials following the applicable budget process and considering public priorities
- The vendor that selected the algorithm
- The department with the largest dataset