Time Series Analysis & Forecasting — Quiz | Data Science using Python

Question 1

Why can't you use a random train/test split on a time series?

A It is too slow

B Order carries information; you must train on the past and test on the future

C It uses too much memory

D Random splits are only for images

Question 2

What three components does seasonal decomposition separate?

A Mean, median, mode

B Trend, seasonality, residual

C Train, validation, test

D Lag, lead, level

Question 3

In the ADF stationarity test, a p-value below 0.05 means…

A the series is non-stationary

B the series is stationary

C the data is missing

D there is no trend test

Question 4

What does differencing a series do?

A Adds seasonality

B Removes trend to make the series stationary

C Doubles the values

D Deletes outliers

Question 5

In ARIMA(p, d, q), what does d represent?

A The number of autoregressive terms

B The degree of differencing

C The moving-average terms

D The forecast horizon

Question 6

What does exponential smoothing (Holt-Winters) emphasise?

A The oldest observations

B Recent observations more heavily, with trend and seasonality

C Only the mean

D Random values

Question 7

When engineering lag features for ML forecasting, what must you avoid?

A Using past values

B Letting a feature include information from the time you are predicting (leakage)

C Adding a month column

D Using gradient boosting

Question 8

Which metric expresses forecast error as a percentage?

A MAE

B RMSE

C MAPE

D R²

Question 9

What is a naive forecast baseline?

A A deep neural network

B Predicting the next value equals the last (or last season's) value

C The average of all data

D A random guess

Question 10

What does backtesting with a rolling origin simulate?

A Shuffling the data

B Repeatedly forecasting forward through history to test stability

C Training on the test set

D Removing seasonality

Question 11

What does Prophet expect its input columns to be named?

A x and y

B date and value

C ds and y

D time and sales

Question 12

Why should a forecast include an uncertainty band?

A To look complex

B Because a single line is overconfident; decision-makers need the likely range

C It is required by pandas

D To hide errors