Book Image

Deep Learning for Time Series Cookbook

By : Vitor Cerqueira, Luís Roque
Book Image

Deep Learning for Time Series Cookbook

By: Vitor Cerqueira, Luís Roque

Overview of this book

Most organizations exhibit a time-dependent structure in their processes, including fields such as finance. By leveraging time series analysis and forecasting, these organizations can make informed decisions and optimize their performance. Accurate forecasts help reduce uncertainty and enable better planning of operations. Unlike traditional approaches to forecasting, deep learning can process large amounts of data and help derive complex patterns. Despite its increasing relevance, getting the most out of deep learning requires significant technical expertise. This book guides you through applying deep learning to time series data with the help of easy-to-follow code recipes. You’ll cover time series problems, such as forecasting, anomaly detection, and classification. This deep learning book will also show you how to solve these problems using different deep neural network architectures, including convolutional neural networks (CNNs) or transformers. As you progress, you’ll use PyTorch, a popular deep learning framework based on Python to build production-ready prediction solutions. By the end of this book, you'll have learned how to solve different time series tasks with deep learning using the PyTorch ecosystem.

Related Content you might be interested in

Current Title:

Small book image
Deep Learning for Time Series Cookbook
Vitor Cerqueira | Luís Roque
Mar, 2024 | 274 pages
Small book image
Azure Strategy and Implementation Guide, Fourth Edition
Jason Milgram | David Rendoacuten | Aaditya Pokkunuri | Greg Leonardo | Jack Lee
May, 2021 | 228 pages
Small book image
Azure Integration Guide for Business
Meacutelony Qin | Joshua Garverick | Trevoir Williams | Jack Lee
Sep, 2023 | 268 pages
Small book image
SAP on Azure Implementation Guide
Nick Morgan | Bartosz Jarkowski
Feb, 2020 | 242 pages
Table of Contents (12 chapters)
Preface
Preface
Who this book is for
What this book covers
To get the most out of this book
Conventions used
Sections
Get in touch
Share Your Thoughts
Download a free PDF copy of this book
Free Chapter
1
Chapter 1: Getting Started with Time Series
Chapter 1: Getting Started with Time Series
Technical requirements
Loading a time series using pandas
Visualizing a time series
Resampling a time series
Dealing with missing values
Decomposing a time series
Computing autocorrelation
Detecting stationarity
Dealing with heteroskedasticity
Loading and visualizing a multivariate time series
Resampling a multivariate time series
Analyzing correlation among pairs of variables
2
Chapter 2: Getting Started with PyTorch
Chapter 2: Getting Started with PyTorch
Technical requirements
Installing PyTorch
Basic operations in PyTorch
Advanced operations in PyTorch
Building a simple neural network with PyTorch
Training a feedforward neural network
Training a recurrent neural network
Training an LSTM neural network
Training a convolutional neural network
3
Chapter 3: Univariate Time Series Forecasting
Chapter 3: Univariate Time Series Forecasting
Technical requirements
Building simple forecasting models
Univariate forecasting with ARIMA
Preparing a time series for supervised learning
Univariate forecasting with a feedforward neural network
Univariate forecasting with an LSTM
Univariate forecasting with a GRU
Univariate forecasting with a Stacked LSTM
Combining an LSTM with multiple fully connected layers
Univariate forecasting with a CNN
Handling trend – taking first differences
Handling seasonality – seasonal dummies and Fourier series
Handling seasonality – seasonal differencing
Handling seasonality – seasonal decomposition
Handling non-constant variance – log transformation
4
Chapter 4: Forecasting with PyTorch Lightning
Chapter 4: Forecasting with PyTorch Lightning
Technical requirements
Preparing a multivariate time series for supervised learning
Training a linear regression model for forecasting with a multivariate time series
Feedforward neural networks for multivariate time series forecasting
LSTM neural networks for multivariate time series forecasting
Monitoring the training process using Tensorboard
Evaluating deep neural networks for forecasting
Using callbacks – EarlyStopping
5
Chapter 5: Global Forecasting Models
Chapter 5: Global Forecasting Models
Technical requirements
Multi-step forecasting with multivariate time series
Multi-step and multi-output forecasting with multivariate time series
Preparing multiple time series for a global model
Training a global LSTM with multiple time series
Global forecasting models for seasonal time series
Hyperparameter optimization using Ray Tune
6
Chapter 6: Advanced Deep Learning Architectures for Time Series Forecasting
Chapter 6: Advanced Deep Learning Architectures for Time Series Forecasting
Technical requirements
Interpretable forecasting with N-BEATS
Optimizing the learning rate with PyTorch Forecasting
Getting started with GluonTS
Training a DeepAR model with GluonTS
Training a Transformer model with NeuralForecast
Training a Temporal Fusion Transformer with GluonTS
Training an Informer model with NeuralForecast
Comparing different Transformers with NeuralForecast
7
Chapter 7: Probabilistic Time Series Forecasting
Chapter 7: Probabilistic Time Series Forecasting
Technical requirements
Introduction to exceedance probability forecasting
Exceedance probability forecasting with an LSTM
Creating prediction intervals using conformal prediction
Probabilistic forecasting with an LSTM
Probabilistic forecasting with DeepAR
Introduction to Gaussian Processes
Using Prophet for probabilistic forecasting
8
Chapter 8: Deep Learning for Time Series Classification
Chapter 8: Deep Learning for Time Series Classification
Technical requirements
Tackling TSC with K-nearest neighbors
Building a DataModule class for TSC
Convolutional neural networks for TSC
ResNets for TSC
Tackling TSC problems with sktime
9
Chapter 9: Deep Learning for Time Series Anomaly Detection
Chapter 9: Deep Learning for Time Series Anomaly Detection
Technical requirements
Time series anomaly detection with ARIMA
Prediction-based anomaly detection using DL
Anomaly detection using an LSTM AE
Building an AE using PyOD
Creating a VAE for time series anomaly detection
Using GANs for time series anomaly detection
10
Index
Index
Why subscribe?
11
Other Books You May Enjoy
Other Books You May Enjoy
Packt is searching for authors like you
Share Your Thoughts
Download a free PDF copy of this book

Detecting stationarity

Stationarity is a central concept in time series analysis and an important assumption made by many time series models. This recipe walks you through the process of testing a time series for stationarity.

Getting ready

A time series is stationary if its statistical properties do not change. It does not mean that the series does not change over time, just that the way it changes does not itself change over time. This includes the level of the time series, which is constant under stationary conditions. Time series patterns such as trend or seasonality break stationarity. Therefore, it may help to deal with these issues before modeling. As we described in the Decomposing a time series recipe, there is evidence that removing seasonality improves the forecasts of deep learning models.

We can stabilize the mean level of the time series by differencing. Differencing is the process of taking the difference between consecutive observations. This process works in two steps:

  1. Estimate the number of differencing steps required for stationarity.
  2. Apply the required number of differencing operations.

How to do it…

We can estimate the required differencing steps with statistical tests, such as the augmented Dickey-Fuller test, or the KPSS test. These are implemented in the ndiffs() function, which is available in the pmdarima library:

from pmdarima.arima import ndiffs
ndiffs(x=series_daily, test='adf')

Besides the time series, we pass test='adf' as an input to set the method to the augmented Dickey-Fuller test. The output of this function is the number of differencing steps, which in this case is 1. Then, we can differentiate the time series using the diff() method:

series_changes = series_daily.diff()

Differencing can also be applied over seasonal periods. In such cases, seasonal differencing involves computing the difference between consecutive observations of the same seasonal period:

from pmdarima.arima import nsdiffs
nsdiffs(x=series_changes, test='ch', m=365)

Besides the data and the test (ch for Canova-Hansen), we also specify the number of periods. In this case, this parameter is set to 365 (number of days in a year).

How it works…

The differenced time series is shown in the following figure.

Figure 1.5: Sample of the series of changes between consecutive periods after differencing

Figure 1.5: Sample of the series of changes between consecutive periods after differencing

Differencing works as a preprocessing step. First, the time series is differenced until it becomes stationary. Then, a forecasting model is created based on the differenced time series. The forecasts provided by the model can be transformed to the original scale by reverting the differencing operations.

There’s more…

In this recipe, we focused on two particular methods for testing stationarity. You can check other options in the function documentation: https://alkaline-ml.com/pmdarima/modules/generated/pmdarima.arima.ndiffs.html.

Previous Section
End of Section 9
Next Section