引導式測驗 · 中等 等級
先知預測測驗
測試 Prophet 趨勢、季節性和假期、加法效應與乘法效應、資料節奏、未來差距、滾動驗證和基準。
問題 1 的 8
哪些部分是指南中描述的先知預測的核心?
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。
- 哪些部分是指南中描述的先知預測的核心?
- Trend, recurring seasonal effects and optional holidays
- Only the last observed value and a random shock
- A clustering label for every date and no trend
- Only a holiday table with no time-series history
- 先知的官方指南什麼時候說該方法最有效?
- When no past observations are available
- When strong seasonal effects and several seasons of history are present
- When every observation occurs at an irregular, unknown time
- When the future must match the past exactly
- 隨著商店整體銷售水準的成長,季節性波動也會增加。哪一種先知設定可能比較適合這種模式?
- A constant additive seasonal amount for every level
- Multiplicative seasonality, checked against later data
- Removing all seasonal terms from the model
- Forecasting only dates outside the observed cadence
- 必須提供什麼才能獲得指定的節日效果以告知未來的先知日期?
- Relevant holiday dates in the historical and forecast calendar
- Only a description of the holiday in the model prompt
- A random training-test split
- A guarantee that the holiday will have the same impact each year
- 為什麼僅包含工作日的歷史記錄對於未觀察到的周末時段的預測效果不佳?
- Prophet automatically knows every local weekend behavior
- The data must be converted to monthly observations first
- Seasonal positions absent from training are not well identified
- Weekly seasonality is always zero on Saturday and Sunday
- A team has monthly observations. What forecast cadence does the guide recommend first?
- Only annual totals regardless of the user's task
- Daily values, because the software can produce them
- Sub-hourly values to increase the sample size
- Future months, avoiding unsupported daily interpolation between observed dates
- How should Prophet be evaluated for a future forecasting task?
- Shuffle all dates so later values influence earlier training
- Use only the in-sample fit shown on the training plot
- Choose the best settings after inspecting the final holdout period
- Use time-ordered cutoffs and compare with a suitable simple baseline
- What limitation remains when Prophet displays a narrow prediction interval?
- It proves no future disruption can occur
- It makes the holiday calendar unnecessary
- Unmodeled shocks or regime changes may fall outside the fitted assumptions
- The point forecast must be exactly right