DSE Time Series Analysis and Forecasting 3

Question 1

What capability does a SARIMA model add over a standard ARIMA model?

Accepted Answer

Explicitly capturing repeating seasonal patterns

Answer

Handling non-stationary series with structural breaks

Answer

Modeling multiple correlated time series simultaneously

Answer

Learning long-range dependencies spanning decades

Question 2

Why is walk-forward (time series cross-validation) preferred over standard k-fold cross-validation for time series data?

Accepted Answer

It respects temporal ordering and prevents future data from leaking into the training set

Answer

It requires substantially less computational power to run

Answer

It automatically selects optimal ARIMA parameters

Answer

It handles non-stationary data without differencing

Question 3

Facebook Prophet is best described as which type of forecasting tool?

Accepted Answer

A decomposable additive model that handles trends, seasonality, and holiday effects automatically

Answer

A deep learning model for image-based time series classification

Answer

An LSTM variant specifically optimized for financial time series

Answer

A statistical test for detecting structural change points in time series

Question 4

In time series analysis, what does 'cointegration' describe?

Accepted Answer

A long-run equilibrium relationship between two or more individually non-stationary series

Answer

Two time series that are both stationary and have zero cross-correlation

Answer

The process of applying differencing twice (d=2) to achieve stationarity

Answer

A method for combining multiple seasonal decomposition models

Question 5

Why are Long Short-Term Memory (LSTM) networks preferred over standard RNNs for long time series sequences?

Accepted Answer

LSTMs use gating mechanisms to solve the vanishing gradient problem, enabling learning of long-range dependencies

Answer

LSTMs require significantly less training data than standard RNNs

Answer

LSTMs can only process univariate sequences, making them simpler to train

Answer

LSTMs automatically remove seasonality from data before training begins

Question 6

What does Mean Absolute Percentage Error (MAPE) express about forecast accuracy?

Accepted Answer

Forecast errors expressed as a percentage of actual values, making accuracy scale-independent

Answer

The average absolute difference in raw units between predicted and actual values

Answer

The Pearson correlation between forecasts and actual observations

Answer

The variance of forecast errors measured across the entire time horizon

Question 7

Which is a well-known limitation of using MAPE as a time series forecast accuracy metric?

Accepted Answer

MAPE is undefined when actual values equal zero and penalizes over-forecasts more than under-forecasts

Answer

MAPE cannot be computed for daily or weekly forecast horizons

Answer

MAPE is only valid for stationary time series without trends

Answer

MAPE requires the forecast horizon to be fixed before model training