Specialized Models: Time Series and Survival Analysis Quiz Answers

Specialized Models: Time Series and Survival Analysis Week 01 Quiz Answers

Quiz 01: Check for Understanding

Q1. This is the concept that best defines a Time Series:

• A sequence of data points organized in time order.
• A sequence that captures data sporadically.
• Data collected irregularly.
• Data suited for predictive modeling.

Q2. Which of the following modeling challenges is common to BOTH time series forecasting and standard linear regression?

• Data may be missing or unavailable.
• The target and feature variables are the same.
• Available observations are not independent of each other.
• Testing samples are fundamentally different from training samples.

Q3. Which of the following assumptions is generally required for time series analysis?

• Observed data are independent over time.
• The process generating the data remains the same over time.
• The expected value of the time series one period ahead is constant.
• For any time period, the observations are normally-distributed with a constant variance.

Quiz 02: Check for Understanding

Q1. This time series component is related to the long term direction of the series:

• Trend
• Seasonality
• Residual
• Confidence interval

Q2. This time series component is related to the periodic behavior of the time series:

• Trend
• Seasonality
• Residual
• Confidence interval

Q3. This time series component is related to the irregular fluctuations of the time series:

• Trend
• Seasonality
• Residual
• Confidence interval

End of Module Quiz

Q1. This time series component is related to the long term direction of the series:

• Trend
• Seasonality
• Residual
• Confidence interval

Q2. This is the type of decomposition model that is used if the magnitudes of the seasonal and residual values fluctuate with trend:

• Additive Decomposition Model
• Multiplicative Decomposition Model
• Pseudo-additive Decomposition Model
• Seasonal-Residual Decomposition Model

Q3. This decomposition model assumes that the seasonal and residual magnitudes are independent of trend.

• Additive Decomposition Model
• Multiplicative Decomposition Model
• Pseudo-additive Decomposition Model
• Seasonal-Residual Decomposition Model

Q4. These are all Time Series decomposition techniques, except:

• Exponential smoothing
• Frequency-based methods
• Ordinary Least Square models
• Locally Estimated Scatterplot Smoothing (LOESS)

Specialized Models: Time Series and Survival Analysis Week 02 Quiz Answers

Quiz 01: Check for Understanding

Q1. Select the correct statement. When comparing a stationary and a non-stationary Time Series:

• A stationary series has the same mean and variance over time, making a non-stationary series easier to model.
• A stationary series has the same mean and variance over time, making a non-stationary series harder to model.
• A stationary series has increasing mean and variance over time, making a non-stationary series easier to model.
• A stationary series has increasing mean and variance over time, making a non-stationary series harder to model.

Q2. Which of the following statements about time series analysis is TRUE?

• Over a long enough time period, most observed processes are stationary.
• Data preprocessing and feature engineering are not applicable to time series analysis.
• Predicting future values requires assumptions about how conditions in the future will change.
• A non-stationary time series can be transformed and modeled as a stationary timeseries.

Quiz 02: Check for Understanding

Q1. Which of the following smoothing techniques is appropriate for data with a trend but no seasonality?

• Simple (average) smoothing
• Single Exponential Smoothing
• Double Exponential Smoothing
• Triple Exponential Smoothing

Q2. Which of the following smoothing techniques is appropriate for data with both trend and seasonality?

• Simple (average) smoothing
• Single Exponential Smoothing
• Double Exponential Smoothing
• Triple Exponential Smoothing

Q3. (True/False) Smoothing techniques require scaled data to be effective.

• True
• False

End of Module Quiz

Q1. (True/False) An important component of time series modeling involves plotting and visual inspection of the data.

• True
• False

Q2. (True/False) If forecast residuals approximate white noise, the model is likely a bad fit and a different model should be selected.

• True
• False

Q3. This smoothing model assumes that the seasonal and residual magnitudes are independent of trend.

Additive Smoothing Model

Multiplicative Smoothing Model

Pseudo-additive Smoothing Model

Seasonal-Residual Smoothing Model

Q4. These are all Time Series smoothing techniques, except:

• Exponential smoothing
• Frequency-based methods
• Ordinary Least Square models
• Locally Estimated Scatterplot Smoothing (LOESS)

Q5. (True/False) A common limitation of forecasts based on smoothed data is their high sensitivity to outliers and unusual historical values.

• True
• False

Specialized Models: Time Series and Survival Analysis Week 03 Quiz Answers

Quiz 01: Check for Understanding

Q1. (True/False) ARMA models combine two models.

• True
• False

Q2. This term can be used interchangeably with the term ARMA model:

• ARIMA model
• Autoregressive model
• Box-Jenkins approach
• Moving average model

Q3. These are all facts about ARMA models, except:

• The Time Series they model should not contain a seasonal component
• The Time Series they model is assumed to be stationary.
• A good rule of thumb is to have at least 1000 observations to fit.
• There are three stages in building an ARMA model: identification, estimation, and evaluation.

End of Module Quiz

Q1. Which of the following is a characteristic of an autoregressive (AR) model?

• Observations are assumed to be uncorrelated over time.
• A fixed number of past forecast errors are used to predict future values.
• A fixed number of past forecast values are used to predict future values.
• The number of historical values used to predict future values increases over time.

Q2. Which of the following is a characteristic of a moving average (MA) model?

• Observations are assumed to be uncorrelated over time.
• A fixed number of past forecast values are used to predict future values.
• A fixed number of past forecast errors are used to predict future values.
• The number of historical values used to predict future values increases over time.

Q3. An ARIMA model without differencing (I=0) is equivalent to which of the following approaches?

• Combining several autoregressive (AR) models with different lags.
• Combining several moving average (MA) models with different lags.
• The sum of an autoregressive (AR) and moving average (MA) model.
• An autoregressive model in first differences (I=1).

Q4. These are all ways to detect seasonality, except:

• Stationarity tests
• Autocorrelation plots
• Seasonal subseries plots
• Prior knowledge about holidays or seasonal effects in the data

Q5. This plot summarizes the 2-way correlation between a variable and its past values:

• Stationarity plot
• Autocorrelation plot
• Seasonal subseries plot
• White noise plot

Specialized Models: Time Series and Survival Analysis Week 04 Quiz Answers

Quiz 01: Check for Understanding

Q1. These are all advantages of Deep Learning approaches to Time Series forecasting, except:

• Automatically learn how to incorporate series characteristics like trend, seasonality, and autocorrelation into predictions.
• Able to capture very complex patterns.
• Tend to perform best with large training datasets.
• Can simultaneously model many related series instead of treating each separately.

Q2. This Deep Learning approach is well suited for long input sequences as is often the case in Time series in which sequences are often hundreds of steps:

• Modular Neural Network
• Long-Short Term Memory
• Recurrent Neural Networks
• Convolutional Neural Networks

Q3. (True/False) Comparing LSTMs vs RNNs, LSTMs tend to take longer to train (slower backpropagation) and can be more prone to overfitting.

• True
• False

Q4. These are all best practices to select a Deep Learning approach for a Forecasting task, EXCEPT:

• If sequences are many time steps long, an RNN may perform poorly.
• If training time is an issue, using a LSTM may be too cumbersome.
• Graphics processing units (GPUs) speed up all neural network training, but are especially recommended when training LSTMs on large datasets.
• A common approach is to use an ensemble of both and LSTM and an RNN to improve the forecast.

Quiz 02: Check for Understanding

Q1. (True/False) In the use case of predicting customer churn, you can use Survival Analysis to calculate the expected time until churn.

• True
• False

Q2. (True/False) You can use the Keplen Meier curve to create a Survival Analysis model

• True
• False

Q3. This Survival Analysis model assumes that the proportional impact on the hazard rate doesn’t change over time:

• Survival random forests
• Parametric survival models
• Cox proportional hazards regression
• Survival trees

End of Module Quiz

Q1. For which of the customer churn problems below is survival analysis MOST DIRECTLY suitable?

• Estimating the length of time until a customer will churn
• Predicting the likelihood that an individual customer will churn
• Estimating which groups of customers are most likely to churn
• Determining which factors are most important for keeping a customer from churning

Q2. Survival analysis is MOST useful for modeling:

• any machine learning problem with a time component.
• the time until a binary event occurs, with censored data.
• the seasonality component of non-integrated time series data.
• single-variable machine learning problems with a time component.

Q3. (True/False) An example of censoring involves data used to estimate the length of time a customer will remain with the company if it includes individuals who have not yet left.

• True
• False

Q4. (True/False) Both Logistic Regression and Survival Analysis are appropriate for predicting censored target variables.

• True
• False

Q5. Which of the following is an advantage of using Deep Learning approaches (relative to standard approaches) to model time series?

• It is often easier to model multiple series jointly using Deep Learning models.
• Deep Learning approaches allow forecasting of series that are nonstationary.
• Deep Learning algorithms allow for easier interpretation of underlying patterns.
• Neural network models use more parameters which makes them more flexible.

Get All Course Quiz Answers of IBM Machine Learning Professional Certificate

Exploratory Data Analysis for Machine Learning Quiz Answers

Supervised Machine Learning: Regression Quiz Answers

Supervised Machine Learning: Classification Coursera Quiz Answers

Unsupervised Machine Learning Coursera Quiz Answers

Deep Learning and Reinforcement Learning Quiz Answers

Specialized Models: Time Series and Survival Analysis Quiz Answers