Time Series Q&A
20 Core Questions
Interview Prep
Time Series & Forecasting: Interview Q&A
Short questions and answers on time series fundamentals: trend, seasonality, stationarity and common forecasting approaches.
Trend
Seasonality
Stationarity
Lags
1
What is a time series?
⚡ Beginner
Answer: A time series is a sequence of observations ordered in time, such as daily sales, sensor readings or stock prices.
2
What are trend and seasonality?
⚡ Beginner
Answer: Trend is a long-term increase or decrease; seasonality is a repeating pattern at fixed periods (daily, weekly, yearly).
3
What does stationarity mean in time series?
📊 Intermediate
Answer: A stationary series has constant mean, variance and autocovariance over time.
4
Why is stationarity often important for classical models like ARIMA?
📊 Intermediate
Answer: Many theoretical results and parameter estimation methods for ARIMA assume stationarity to be valid.
5
What is differencing and when is it used?
📊 Intermediate
Answer: Differencing subtracts the previous value \((y_t - y_{t-1})\) to remove trend or make a series more stationary.
6
What is autocorrelation in a time series?
📊 Intermediate
Answer: Autocorrelation measures how strongly the series is correlated with lagged versions of itself.
7
What is the basic idea behind ARIMA models?
🔥 Advanced
Answer: ARIMA combines autoregressive (AR), differencing (I) and moving average (MA) components to model time series.
8
How can you create features for time series using ML models like trees?
📊 Intermediate
Answer: Common features: lags, rolling means, rolling std, calendar features (day of week, month, holidays).
9
What is the difference between one-step-ahead and multi-step forecasting?
📊 Intermediate
Answer: One-step forecasts the next time point; multi-step forecasts several future steps (possibly by iterating one-step or direct models).
10
Why is data leakage a special concern in time series?
🔥 Advanced
Answer: Using future information to predict the past (e.g., random train/test splits) gives unrealistically good performance estimates.
11
How should you split time series data for validation?
📊 Intermediate
Answer: Use time-based splits (train on earlier periods, validate on later periods) or rolling/expanding window validation.
12
Name some common error metrics for forecasting.
⚡ Beginner
Answer: Metrics: MAE, RMSE, MAPE, sMAPE, sometimes custom business metrics.
13
When would you use simple baselines like naive or seasonal naive forecasts?
⚡ Beginner
Answer: Always as a reference point; complex models should beat baselines that simply repeat the last value or last season’s value.
14
What is concept drift in time series/forecasting?
🔥 Advanced
Answer: Concept drift occurs when the underlying data-generating process changes over time, making old models less accurate.
15
How do you handle missing values in time series?
📊 Intermediate
Answer: Methods include forward/backward fill, interpolation, model-based imputation, or special indicators for missing events.
16
Give some real-world applications of time series forecasting.
⚡ Beginner
Answer: Applications: demand forecasting, energy load prediction, financial markets, traffic prediction.
17
How can neural networks be used for time series forecasting?
🔥 Advanced
Answer: Using RNNs/LSTMs, 1D CNNs, transformers or generic deep nets on lag features and covariates.
18
Why is it important to align features to the forecast horizon correctly?
🔥 Advanced
Answer: You must avoid using information from the future when predicting earlier time points, or metrics will be overly optimistic.
19
When would a simple model be preferable to a complex one for time series?
📊 Intermediate
Answer: When data is short, noisy, or patterns are simple, simple models are easier to maintain and often more robust.
20
What is the key message to remember about time series forecasting?
⚡ Beginner
Answer: Understand the temporal structure, leakage risks and business use-case first—good forecasting is as much about setup as about the chosen model.
Quick Recap: Time Series
If you can talk clearly about trend, seasonality, stationarity and leakage, you’re in a strong position for most time-series interview questions.