Gids-gekoppelde quiz · Middelmatig Niveau
AI voor FP&A en quiz over financiële prognoses
Test op drivers gebaseerde versus machine learning-voorspellingen, rolling forecasts, backtesting en lekkage, en de grenzen van door AI opgestelde variantiecommentaar.
Vraag 1 van 8
Wat maakt een prognose in FP&A gebaseerd op drivers?
Bekijk elke vraag8 vragen
Antwoorden en uitleg verschijnen terwijl je de quiz doet.
- Wat maakt een prognose in FP&A gebaseerd op drivers?
- It is produced only by a neural network
- Financial lines are linked to operating drivers such as headcount, units, price or win rate
- It uses last year's actuals with no changes
- It is prepared only by the board of directors
- Wat definieert een voortschrijdende prognose zoals beschreven in de handleiding?
- A forecast made once per year and never revised
- A forecast only of cash balances
- A forecast that extends a set number of months ahead, such as 12 or 18, and is updated regularly
- A forecast that covers only the current month
- Waarom zou een team een machine learning-voorspelling backtesten aan de hand van een eenvoudige basislijn, zoals dezelfde week-vorig jaar?
- On short, noisy or disrupted histories, simple baselines can beat complex models
- Simple baselines are required by accounting standards
- Machine learning models cannot be backtested
- Baselines are always more accurate
- Wat is een voorbeeld van gegevenslekken in een backtest voor prognoses?
- Using several years of history to train the model
- Reconciling SKU forecasts up to the total
- Publishing prediction intervals
- Including actual promotion results that would not have been known when the forecast was made
- Wat is het grootste risico als je een taalmodel variantiecommentaar laat opstellen zonder extra input?
- It cannot read tables
- It may state plausible but invented causes, since the data shows the size of a variance but not the reason
- It always refuses to discuss budgets
- It converts all figures into a different currency
- What does forecast value added (FVA) analysis measure?
- Whether human overrides improve accuracy compared with the statistical baseline
- The dollar value of the forecast's revenue
- How long a forecast takes to produce
- The number of drivers in a model
- In forecast evaluation, what does bias refer to?
- The model's preference for one software vendor
- The random scatter of errors around zero
- A consistent tendency to over- or under-forecast
- The difference between two currencies
- In which situation does the guide suggest statistical and machine learning forecasts are most likely to struggle?
- Forecasting many stable product lines with long histories
- Producing baselines for thousands of cost centers
- Updating a monthly rolling forecast
- New products, pricing changes or economic shocks the model has never seen