Forecasting Time Series Data with Prophet: Build, improve, and optimize time series forecasting models using Meta's advanced forecasting tool, 2nd Edition

Description

Create and improve fully automated forecasts for time series data with strong seasonal effects, holidays, and additional regressors using Python Purchase of the print or Kindle book includes a free PDF eBook Key FeaturesExplore Prophet, the open source forecasting tool developed at Meta, to improve your forecastsCreate a forecast and run diagnostics to understand forecast qualityFine-tune models to achieve high performance and report this performance with concrete statisticsBook Description Prophet empowers Python and R developers to build scalable time series forecasts. This book will help you to implement Prophet's cutting-edge forecasting techniques to model future data with high accuracy using only a few lines of code. You'll begin by exploring the evolution of time series forecasting, from basic early models to present-day advanced models. After the initial installation and setup, you'll take a deep dive into the mathematics and theory behind Prophet. You'll then cover advanced features such as visualizing your forecasts, adding holidays and trend changepoints, and handling outliers. You'll use the Fourier series to model seasonality, learn how to choose between an additive and multiplicative model, and understand when to modify each model parameter. This updated edition has a new section on modeling shocks such as COVID. Later on in the book you'll see how to optimize more complicated models with hyperparameter tuning and by adding additional regressors to the model. Finally, you'll learn how to run diagnostics to evaluate the performance of your models and discover useful features when running Prophet in production environments. By the end of this book, you'll be able to take a raw time series dataset and build advanced and accurate forecasting models with concise, understandable, and repeatable code. What you will learnUnderstand the mathematics behind Prophet's modelsBuild practical forecasting models from real datasets using PythonUnderstand the different modes of growth that time series often exhibitDiscover how to identify and deal with outliers in time series dataFind out how to control uncertainty intervals to provide percent confidence in your forecastsProductionalize your Prophet models to scale your work faster and more efficientlyWho this book is for This book is for business managers, data scientists, data analysts, machine learning engineers, and software engineers who want to build time-series forecasts in Python or R. To get the most out of this book, you should have a basic understanding of time series data and be able to differentiate it from other types of data. Basic knowledge of forecasting techniques is a plus. Table of ContentsThe History and Development of Time Series ForecastingGetting Started with ProphetHow Prophet WorksHandling Non-Daily DataWorking with SeasonalityForecasting Holiday EffectsControlling Growth ModesInfluencing Trend ChangepointsIncluding Additional RegressorsAccounting for Outliers and Special EventsManaging Uncertainty IntervalsPerforming Cross-ValidationEvaluating Performance MetricsProductionalizing Prophet Cover Title Page Copyright Contributors Table of Contents Preface Part 1: Getting Started with Prophet Chapter 1: The History and Development of Time Series Forecasting Understanding time series forecasting The problem with dependent data Moving averages and exponential smoothing ARIMA ARCH/GARCH Neural networks Prophet Recent developments NeuralProphet Google’s “robust time series forecasting at scale” LinkedIn’s Silverkite/Greykite Uber’s Orbit Summary Chapter 2: Getting Started with Prophet Technical requirements Installing Prophet Installation on macOS Installation on Windows Installation on Linux Building a simple model in Prophet Interpreting the forecast DataFrame Understanding components plots Summary Chapter 3: How Prophet Works Technical requirements Facebook’s motivation for building Prophet Analyst-in-the-loop forecasting The math behind Prophet Linear growth Logistic growth Seasonality Holidays Summary Part 2: Seasonality, Tuning, and Advanced Features Chapter 4: Handling Non-Daily Data Technical requirements Using monthly data Using sub-daily data Using data with regular gaps Summary Chapter 5: Working with Seasonality Technical requirements Understanding additive versus multiplicative seasonality Controlling seasonality with the Fourier order Adding custom seasonalities Adding conditional seasonalities Regularizing seasonality Global seasonality regularization Local seasonality regularization Summary Chapter 6: Forecasting Holiday Effects Technical requirements Adding default country holidays Adding default state/province holidays Creating custom holidays Creating multi-day holidays Regularizing holidays Global holiday regularization Individual holiday regularization Summary Chapter 7: Controlling Growth Modes Technical requirements Applying linear growth Understanding the logistic function Saturating forecasts Increasing logistic growth Non-constant cap Decreasing logistic growth Applying flat growth Creating a custom trend Summary Chapter 8: Influencing Trend Changepoints Technical requirements Automatic trend changepoint detection Default changepoint detection Regularizing changepoints Specifying custom changepoint locations Summary Chapter 9: Including Additional Regressors Technical requirements Adding binary regressors Adding continuous regressors Interpreting the regressor coefficients Summary Chapter 10: Accounting for Outliers and Special Events Technical requirements Correcting outliers that cause seasonality swings Correcting outliers that cause wide uncertainty intervals Detecting outliers automatically Winsorizing Standard deviation The moving average Error standard deviation Modeling outliers as special events Modeling shocks such as COVID-19 lockdowns Summary Chapter 11: Managing Uncertainty Intervals Technical requirements Modeling uncertainty in trends Modeling uncertainty in seasonality Summary Part 3: Diagnostics and Evaluation Chapter 12: Performing Cross-Validation Technical requirements Performing k-fold cross-validation Performing forward-chaining cross-validation Creating the Prophet cross-validation DataFrame Parallelizing cross-validation Summary Chapter 13: Evaluating Performance Metrics Technical requirements Understanding Prophet’s metrics Mean squared error Root mean squared error Mean absolute error Mean absolute percent error Median absolute percent error Symmetric mean absolute percent error Coverage Choosing the best metric Creating a Prophet performance metrics DataFrame Handling irregular cut-offs Tuning hyperparameters with grid search Summary Chapter 14: Productionalizing Prophet Technical requirements Saving a model Updating a fitted model Making interactive plots with Plotly Plotly forecast plot Plotly components plot Plotly single component plot Plotly seasonality plot Summary Index About Packt Other Books You May Enjoy