Time-series handling, splitting, and forecasting evaluation: Practice Questions — Advanced Data Structures (NVIDIA-Certified Associate: Accelerated Data Science)
Practice Questions: Time-Series Handling, Splitting, and Forecasting Evaluation This set of multiple-choice questions is designed to help candidates...
Practice Questions: Time-Series Handling, Splitting, and Forecasting Evaluation
This set of multiple-choice questions is designed to help candidates prepare for the NVIDIA-Certified Associate: Accelerated Data Science exam, focusing specifically on time-series data handling, splitting strategies, and forecasting evaluation techniques using GPU-accelerated tools like cuDF.
Which of the following is the best method to split a time-series dataset for training and testing to avoid data leakage?
- A. Randomly shuffle and split the data into training and testing sets
- B. Use the last portion of the time-series as the test set, preserving temporal order
- C. Split the data based on categorical labels
- D. Use cross-validation with random folds
Correct answer: B
Explanation: Time-series data must maintain temporal order to prevent future data from leaking into the training set. Using the last portion as the test set preserves this order.
When handling missing timestamps in a time-series dataset using cuDF, which approach is most appropriate?
- A. Drop all rows with missing timestamps
- B. Use forward fill or interpolation to impute missing values
- C. Replace missing timestamps with zeros
- D. Ignore missing timestamps and proceed with modeling
Correct answer: B
Explanation: Forward fill or interpolation helps maintain continuity in time-series data, which is critical for accurate forecasting and analysis.
Which metric is commonly used to evaluate forecasting accuracy when the scale of the data varies significantly?
- A. Mean Absolute Error (MAE)
- B. Root Mean Squared Error (RMSE)
- C. Mean Absolute Percentage Error (MAPE)
- D. R-squared
Correct answer: C
Explanation: MAPE expresses errors as percentages, making it suitable when comparing forecasts across different scales.
In time-series cross-validation, what is the purpose of using a rolling forecasting origin?
- A. To randomly shuffle data for better generalization
- B. To evaluate the model’s performance on multiple sequential test sets
- C. To increase the size of the training set by duplication
- D. To reduce computational cost by limiting test data
Correct answer: B
Explanation: Rolling forecasting origin allows evaluation over multiple time windows, reflecting how the model performs as new data becomes available.
Which cuDF function is most suitable for resampling irregular time-series data to a regular frequency?
- A. cudf.DataFrame.dropna()
- B. cudf.Series.interpolate()
- C. cudf.DataFrame.resample()
- D. cudf.Series.fillna()
Correct answer: C
Explanation: The resample() function allows aggregation or interpolation of data points to a consistent time interval, essential for regular time-series analysis.
What is the main advantage of using time-based splitting over random splitting in time-series forecasting?
- A. It reduces the size of the dataset
- B. It ensures the model trains on future data
- C. It preserves the temporal sequence to mimic real-world forecasting
- D. It increases model complexity
Correct answer: C
Explanation: Time-based splitting respects the chronological order, which is critical for realistic forecasting scenarios.
Which evaluation approach helps assess model performance stability over different forecast horizons?
- A. Single-step forecasting evaluation
- B. Multi-step rolling forecast evaluation
- C. Random sampling of forecast points
- D. Ignoring forecast horizon and evaluating overall error
Correct answer: B
Explanation: Multi-step rolling forecast evaluation tests the model’s ability to predict multiple future points and its stability over time.
More in this topic
Ready to test your knowledge?
Put what you've learned into practice with a quick quiz and track your progress.
Test your knowledge →