Table of Contents
About the Authors
About the Technical Reviewer
Preface
Chapter 1: Getting Started with Time Series
Recipe 1-1A. Reading Time Series Objects (Air Passengers)
Problem
Solution
How It Works
Step 1A-1. Import the required libraries.
Step 1A-2. Write a parsing function for the datetime column.
Step 1A-3. Read the data.
Recipe 1-1B. Reading Time Series Objects (India GDP Data)
Problem
Solution
How It Works
Step 1B-1. Import the required libraries.
Step 1B-2. Read India’s GDP time series data.
Step 1B-3. Plot the time series.
Step 1B-4. Store and retrieve as a pickle.
Recipe 1-2. Saving Time Series Objects
Problem
Solution
How It Works
Step 2-1. Save the previously loaded time series object.
Recipe 1-3A. Exploring Types of Time Series Data: Univariate
Problem
Solution
How It Works
Step 3A-1. Import the required libraries.
Step 3A-2. Read the time series data.
Step 3A-3. Plot the time series.
Recipe 1-3B. Exploring Types of Time Series Data: Multivariate
Problem
Solution
How It Works
Step 3B-1. Import the required libraries.
Step 3B-2. Write the parsing function.
Step 3B-3. Load the dataset.
Step 3B-4. Do basic preprocessing.
Step 3B-5. Plot each series.
Recipe 1-4A. Time Series Components: Trends
Problem
Solution
How It Works
Step 4A-1. Import the required libraries.
Step 4A-2. Write the parsing function.
Step 4A-3. Load the dataset.
Step 4A-4. Plot the time series.
Recipe 1-4B. Time Series Components: Seasonality
Problem
Solution
How It Works
Step 4B-1. Import the required libraries.
Step 4B-2. Read the data.
Step 4B-3. Plot the time series.
Step 4B-4. Plot a box plot by month.
Step 4B-5. Plot a box plot by year.
Recipe 1-4C. Time Series Components: Seasonality (cont’d.)
Problem
Solution
How It Works
Step 4C-1. Import the required libraries.
Step 4C-2. Read tractor sales data.
Step 4C-3. Set a datetime series to use as an index.
Step 4C-4. Format the data.
Step 4C-5. Plot the time series.
Step 4C-6. Plot a box plot by month.
Recipe 1-5A. Time Series Decomposition: Additive Model
Problem
Solution
How It Works
Step 5A-1. Load the required libraries.
Step 5A-2. Read and process retail turnover data.
Step 5A-3. Plot the time series.
Step 5A-4. Decompose the time series.
Step 5A-5. Separate the components.
Recipe 1-5B. Time Series Decomposition: Multiplicative Model
Problem
Solution
How It Works
Step 5B-1. Load the required libraries.
Step 5B-2. Load air passenger data.
Step 5B-3. Process the data.
Step 5B-4. Decompose the time series.
Step 5B-5. Get the seasonal component.
Recipe 1-6. Visualization of Seasonality
Problem
Solution
How It Works
Step 6-1. Import the required libraries.
Step 6-2. Load the data.
Step 6-3. Process the data.
Step 6-4. Pivot the table.
Step 6-5. Plot the line charts.
Step 6-6. Plot the box plots.
Chapter 2: Statistical Univariate Modeling
Recipe 2-1. Moving Average (MA) Forecast
Problem
Solution
How It Works
Step 1-1. Import the required libraries.
Step 1-2. Read the data.
Step 1-3. Preprocess the data.
Step 1-4. Plot the time series.
Step 1-5. Use a rolling mean to get the moving average.
Step 1-6. Plot the forecast vs. the actual.
Recipe 2-2. Autoregressive (AR) Model
Problem
Solution
How It Works
Step 2-1. Import the required libraries.
Step 2-2. Load and plot the dataset.
Step 2-3. Check for stationarity of the time series data.
Step 2-4. Find the order of the AR model to be trained.
Step 2-5. Create training and test data.
Step 2-6. Call and fit the AR model.
Step 2-7. Output the model summary.
Step 2-8. Get the predictions from the model.
Step 2-9. Plot the predictions vs. actuals.
Recipe 2-3. Autoregressive Moving Average (ARMA) Model
Problem
Solution
How It Works
Step 3-1. Import the required libraries.
Step 3-2. Load the data.
Step 3-3. Preprocess the data.
Step 3-4. Plot the time series.
Step 3-5. Do a train-test split.
Step 3-6. Plot time the series after the train-test split.
Step 3-7. Define the actuals from training.
Step 3-8. Initialize and fit the ARMA model.
Step 3-9. Get the test predictions.
Step 3-10. Plot the train, test, and predictions as a line plot.
Step 3-11. Calculate the RMSE score for the model.
Recipe 2-4. Autoregressive Integrated Moving Average (ARIMA) Model
Problem
Solution
How It Works
Step 4-1. Make the data stationary by differencing.
Step 4-2. Check the ADF (Augmented Dickey-Fuller) test for stationarity.
Step 4-3. Get the Auto Correlation Function and Partial Auto Correlation Function values.
Step 4-4. Plot the ACF and PACF to get p- and q-values.
Step 4-5. Initialize and fit the ARIMA model.
Step 4-6. Get the test predictions.
Step 4-7. Plot the train, test, and predictions as a line plot.
Step 4-8. Calculate the RMSE score for the model.
Recipe 2-5. Grid Search Hyperparameter Tuning for ARIMA Model
Problem
Solution
How It Works
Step 5-1. Write a function to evaluate the ARIMA model.
Step 5-2. Write a function to evaluate multiple models through grid search hyperparameter tuning.
Step 5-3. Perform the grid search hyperparameter tuning by calling the defined functions.
Step 5-4. Initialize and fit the ARIMA model with the best configuration.
Step 5-5. Get the test predictions.
Step 5-6. Plot the train, test, and predictions as a line plot.
Step 5-7. Calculate the RMSE score for the model.
Recipe 2-6. Seasonal Autoregressive Integrated Moving Average (SARIMA) Model
Problem
Solution
How It Works
Step 6-1. Initialize and fit the SARIMA model.
Step 6-2. Get the test predictions.
Step 6-3. Plot the train, test, and predictions as a line plot.
Step 6-4. Calculate the RMSE score for the model.
Recipe 2-7. Simple Exponential Smoothing (SES) Model
Problem
Solution
How It Works
Step 7-1. Initialize and fit the SES model.
Step 7-2. Get the test predictions.
Step 7-3. Plot the train, test, and predictions as a line plot.
Step 7-4. Calculate the RMSE score for the model.
Recipe 2-8. Holt-Winters (HW) Model
Problem
Solution
How It Works
Step 8-1. Initialize and fit the HW model.
Step 8-2. Get the test predictions.
Step 8-3. Plot the train, test, and predictions as a line plot.
Step 8-4. Calculate the RMSE score for the model.
Chapter 3: Advanced Univariate and Statistical Multivariate Modeling
Recipe 3-1. FBProphet Univariate Time Series Modeling
Problem
Solution
How It Works
Step 1-1. Import the required libraries.
Step 1-2. Read the data.
Step 1-3. Create the training dataset.
Step 1-4. Initialize a basic Facebook Prophet model.
Step 1-5. Create the future dataframe for forecasting.
Step 1-6. Getting the predictions.
Step 1-7. Plot the forecast.
Step 1-8. Plot the forecast components.
Recipe 3-2. FBProphet Modeling by Controlling the Change Points
Problem
Solution
How It Works
Step 2-1. Plot the change points.
Step 2-2. Print the change points.
Step 2-3. Check the magnitude of each changepoint.
Step 2-4. Tweak the n_changepoints hyperparameter and forecasting.
Step 2-5. Tweak the changepoint_range hyperparameter and forecasting.
Recipe 3-3. FBProphet Modeling by Adjusting Trends
Problem
Solution
How It Works
Step 3-1. Increase the changepoint_prior_scale hyperparameter.
Step 3-2. Forecast and plot the output.
Step 3-3. Decrease the changepoint_prior_scale hyperparameter.
Step 3-4. Forecast and plot the output.
Recipe 3-4. FBProphet Modeling with Holidays
Problem
Solution
How It Works
Step 4-1. Create a custom holiday dataframe.
Step 4-2. Initialize and fit the Prophet model with the holidays dataframe.
Step 4-3. Create a future dataframe for the forecast.
Step 4-4. Get the forecast.
Recipe 3-5. FBProphet Modeling with Added Regressors
Problem
Solution
How It Works
Step 5-1. Label and encode the type column.
Step 5-2. Get the data in the required format.
Step 5-3. Do a train-test split.
Step 5-4. Initialize the Prophet model and add a regressor.
Step 5-5. Fit the data.
Step 5-6. Forecast the data in the test.
Recipe 3-6. Time Series Forecasting Using Auto-ARIMA
Problem
Solution
How It Works
Step 6-1. Import the required libraries.
Step 6-2. Read the data.
Step 6-3. Preprocess the data.
Step 6-4. Analyze the data pattern.
Step 6-5. Test for stationarity.
Step 6-6. Split the dataset to train and test.
Step 6-7. Build the Auto-ARIMA model.
Step 6-8. Forecast using the test data.
Step 6-9. Evaluate the model.
Recipe 3-7. Multivariate Time Series Forecasting Using the VAR Model
Problem
Solution
How It Works
Step 7-1. Import the required libraries.
Step 7-2. Read the data.
Step 7-3. Preprocess the data.
Step 7-4. Check the stationarity.
Step 7-5. Split the dataset into train-test.
Step 7-6. Build the VAR model and forecast on the test set.
Step 7-7. Evaluate the model.
Chapter 4: Machine Learning Regression–based Forecasting
Recipe 4-1. Formulating Regression Modeling for Time Series Forecasting
Problem
Solution
How It Works
Step 1-1. Install and import the required libraries.
Step 1-2. Collect the data.
Step 1-3. Preprocess the data and create features (feature engineering).
Step 1-4. Select the features.
Step 1-5. Work on the train-test and validation split.
Recipe 4-2. Implementing the XGBoost Model
Problem
Solution
How It Works
Step 2-1. Build the XGBoost model.
Step 2-2. Evaluate the XGBoost model in the test set.
Step 2-3. Evaluate the XGBoost model in the validation set.
Recipe 4-3. Implementing the LightGBM Model
Problem
Solution
How It Works
Step 3-1. Build the LightGBM model.
Step 3-2. Evaluate the LightGBM model in the test set.
Step 3-3. Evaluate the LightGBM model in the validation set.
Recipe 4-4. Implementing the Random Forest Model
Problem
Solution
How It Works
Step 4-1. Build a random forest model.
Step 4-2. Evaluate the LightGBM model in the test set.
Step 4-3. Evaluate the LightGBM model in the validation set.
Recipe 4-5. Selecting the Best Model
Problem
Solution
How It Works
Step 5-1. Evaluate the method.
Step 5-2. Compare performance in the test set.
Step 5-3. Plot the LightGBM model prediction against the actuals in the test set.
Step 5-4. Compare performance in the validation set.
Step 5-5. Plot the LightGBM model prediction against actuals in the validation set.
Chapter 5: Deep Learning–based Time Series Forecasting
Recipe 5-1. Time Series Forecasting Using LSTM
Problem
Solution
How It Works
Step 1-1. Import the required libraries.
Step 1-2. Use DOM_hourly.csv data for analysis.
Step 1-3. Read the data.
Step 1-4. Check for missing data.
Step 1-5. Plot the time series data.
Step 1-6. Write a function to normalize the data.
Step 1-7. Call the normalize_fn function.
Step 1-7. Plot the data after normalization.
Step 1-8. Create a function to perform data preparation and train-test split.
Step 1-9. Call the data_prep function.
Step 1-10. Initialize the LSTM model.
Step 1-11. Create the model summary.
Step 1-12. Fit the model.
Step 1-13. Make the model predictions and print the score.
Step 1-14. Write a function to plot the predictions.
Step 1-15. Call the plotting_actual_vs_pred function.
Recipe 5-2. Multivariate Time Series Forecasting Using the GRU Model
Problem
Solution
How It Works
Step 2-1. Import the required libraries.
Step 2-2. Read the data.
Step 2-3. Analyze the data.
Step 2-4. Preprocess the data.
Step 2-5. Do a train-test split.
Step 2-6. Build the model.
Step 2-7. Evaluate and predict the model.
Recipe 5-3. Time Series Forecasting Using NeuralProphet
Problem
Solution
How It Works
Step 3-1. Import the required libraries.
Step 3-2. Read the data.
Step 3-3. Preprocess the data.
Step 3-4. Build the model and make predictions.
Recipe 5-4. Time Series Forecasting Using RNN
Problem
Solution
How It Works
Step 4-1. Initialize the RNN model.
Step 4-2. Create the model summary.
Step 4-3. Fit the model.
Step 4-4. Make the model predictions and print the score.
Step 4-5 is the same as step 1-14 from Recipe 5-1.
Step 4-6. Call the plotting_actual_vs_pred function.
Index