Codeless Time Series Analysis with KNIME: A practical guide to implementing forecasting models for time series analysis applications

✍ Scribed by Corey Weisinger, Maarit Widmann, Daniele Tonini

• Publisher: Packt Publishing
• Year: 2022
• Tongue: English
• Leaves: 392
• Category: Library

✦ Synopsis

Perform time series analysis using KNIME Analytics Platform, covering both statistical methods and machine learning-based methods

Key Features

• Gain a solid understanding of time series analysis and its applications using KNIME
• Learn how to apply popular statistical and machine learning time series analysis techniques
• Integrate other tools such as Spark, H2O, and Keras with KNIME within the same application

Book Description

This book will take you on a practical journey, teaching you how to implement solutions for many use cases involving time series analysis techniques.

This learning journey is organized in a crescendo of difficulty, starting from the easiest yet effective techniques applied to weather forecasting, then introducing ARIMA and its variations, moving on to machine learning for audio signal classification, training deep learning architectures to predict glucose levels and electrical energy demand, and ending with an approach to anomaly detection in IoT. There's no time series analysis book without a solution for stock price predictions and you'll find this use case at the end of the book, together with a few more demand prediction use cases that rely on the integration of KNIME Analytics Platform and other external tools.

By the end of this time series book, you'll have learned about popular time series analysis techniques and algorithms, KNIME Analytics Platform, its time series extension, and how to apply both to common use cases.

What you will learn

• Install and configure KNIME time series integration
• Implement common preprocessing techniques before analyzing data
• Visualize and display time series data in the form of plots and graphs
• Separate time series data into trends, seasonality, and residuals
• Train and deploy FFNN and LSTM to perform predictive analysis
• Use multivariate analysis by enabling GPU training for neural networks
• Train and deploy an ML-based forecasting model using Spark and H2O

Who this book is for

This book is for data analysts and data scientists who want to develop forecasting applications on time series data. While no coding skills are required thanks to the codeless implementation of the examples, basic knowledge of KNIME Analytics Platform is assumed. The first part of the book targets beginners in time series analysis, and the subsequent parts of the book challenge both beginners as well as advanced users by introducing real-world time series applications.

Table of Contents

1. Introducing Time Series Analysis
2. Introduction to KNIME Analytics Platform
3. Preparing Data for Time Series Analysis
4. Time Series Visualization
5. Time Series Components and Statistical Properties
6. Humidity Forecasting with Classical Methods
7. Forecasting the Temperature with ARIMA and SARIMA Models
8. Audio Signal Classification with an FFT and a Gradient Boosted Forest
9. Training and Deploying a Neural Network to Predict Glucose Levels
10. Predicting Energy Demand with an LSTM Model
11. Anomaly Detection – Predicting Failure with No Failure Examples
12. Predicting Taxi Demand on the Spark Platform
13. GPU Accelerated Model for Multivariate Forecasting
14. Combining KNIME and H2O to Predict Stock Prices

✦ Table of Contents

Cover
Title Page
Copyright and Credits
Dedication
Contributors
Table of Contents
Preface
Part 1: Time Series Basics and KNIME Analytics Platform
Chapter 1: Introducing Time Series Analysis
Understanding TSA
Exploring time series properties and examples
Continuous and discrete time series
Independence and serial correlation
Time series examples
TSA goals and applications
Goals of TSA
Domains of applications and use cases
Exploring time series forecasting techniques
Quantitative forecasting properties and techniques
Summary
Questions
Chapter 2: Introduction to KNIME Analytics Platform
Exploring the KNIME software
Introducing KNIME Analytics Platform for creating data science applications
Introducing KNIME Server for productionizing data science applications
Introducing nodes and workflows
Introducing nodes
Introducing workflows
Searching for and sharing resources on the KNIME Hub
Building your first workflow
Creating a new workflow (group)
Reading and transforming data
Filtering rows
Visualizing data
Building a custom interactive view
Documenting workflows
Configuring the time series integration
Introducing the time series components
Configuring Python in KNIME
Summary
Questions
Chapter 3: Preparing Data for Time Series Analysis
Introducing different sources of time series data
Time granularity and time aggregation
Defining time granularity
Finding the right time granularity
Aggregating time series data
Equal spacing and time alignment
Explaining the concept of equal spacing
Missing value imputation
Defining the different types of missing values
Introducing missing value imputation techniques
Summary
Questions
Chapter 4: Time Series Visualization
Technical requirements
Introducing an energy consumption time series
Describing raw energy consumption data
Clustering energy consumption data
Introducing line plots
Displaying simple dynamics with a line plot
Interpreting the dynamics of a time series based on a line plot
Building a line plot in KNIME
Introducing lag plots
Introducing insights derived from a lag plot
Building a lag plot in KNIME
Introducing seasonal plots
Comparing seasonal patterns in a seasonal plot
Building a seasonal plot in KNIME
Introducing box plots
Inspecting variability of data in a box plot
Building a box plot in KNIME
Summary
Questions
Chapter 5: Time Series Components and Statistical Properties
Technical requirements
Trend and seasonality components
Trend
Seasonality
Decomposition
Autocorrelation
Stationarity
Summary
Questions
Part 2: Building and Deploying a Forecasting Model
Chapter 6: Humidity Forecasting with Classical Methods
Technical requirements
The importance of predicting the weather
Other IoT sensors
The use case
Streaming humidity data from an Arduino sensor
What is an Arduino?
Moving data to KNIME
Storing the data to create a training set
Resampling and granularity
Aligning data timestamps
Missing values
Aggregation techniques
Training and deployment
Types of classic models available in KNIME
Training a model in KNIME
Available deployment options
Building the workflow
Writing model predictions to a database
Summary
Questions
Chapter 7: Forecasting the Temperature with ARIMA and SARIMA Models
Recapping regression
Defining a regression
Introducing the (S)ARIMA models
Requirements of the (S)ARIMA model
How to configure the ARIMA or SARIMA model
Fitting the model and generating forecasts
The data
Summary
Further reading
Questions
Chapter 8: Audio Signal Classification with an FFT and a Gradient-Boosted Forest
Technical requirements
Why do we want to classify a signal?
Windowing your data
Windowing your data in KNIME
What is a transform?
The Fourier transform
Discrete Fourier Transform (DFT)
Fast Fourier Transform (FFT)
Applying the Fourier transform in KNIME
Preparing data for modeling
Reducing dimensionality
Training a Gradient Boosted Forest
Applying the Fourier transform in KNIME
Applying the Gradient Boosted Trees Learner
Deploying a Gradient Boosted Forest
Summary
Questions
Chapter 9: Training and Deploying a Neural Network to Predict Glucose Levels
Technical requirements
Glucose prediction and the glucose dataset
Glucose prediction
The glucose dataset
A quick introduction to neural networks
Artificial neurons and artificial neural networks
The backpropagation algorithm
Other types of neural networks
Training a feedforward neural network to predict glucose levels
KNIME Deep Learning Keras Integration
Building the network
Training the network
Scoring the network and creating the output rule
Deploying an FFNN-based alarm system
Summary
Questions
Chapter 10: Predicting Energy Demand with an LSTM Model
Technical requirements
Introducing recurrent neural networks and LSTMs
Recapping recurrent neural networks
The architecture of the LSTM unit
Forget Gate
Input Gate
Output Gate
Encoding and tensors
Input data
Reshaping the data
Training an LSTM-based neural network
The Keras Network Learner node
Deploying an LSTM network for future prediction
Scoring the forecasts
Summary
Questions
Chapter 11: Anomaly Detection – Predicting Failure with No Failure Examples
Technical requirements
Introducing the problem of anomaly detection in predictive maintenance
Introducing the anomaly detection problem
IoT data preprocessing
Exploring anomalies visually
Detecting anomalies with a control chart
Introducing a control chart
Implementing a control chart
Predicting the next sample in a correctly working system with an auto-regressive model
Introducing an auto-regressive model
Training an auto-regressive model with the linear regression algorithm
Deploying an auto-regressive model
Summary
Questions
Part 3: Forecasting on Mixed Platforms
Chapter 12: Predicting Taxi Demand on the Spark Platform
Technical requirements
Predicting taxi demand in NYC
Connecting to the Spark platform and preparing the data
Introducing the Hadoop ecosystem
Accessing the data and loading it into Spark
Introducing the Spark compatible nodes
Training a random forest model on Spark
Exploring seasonalities via line plots and auto-correlation plot
Preprocessing the data
Training and testing the random forest model on Spark
Building the deployment application
Predicting the trip count in the next hour
Predicting the trip count in the next 24 hours
Summary
Questions
Chapter 13: GPU Accelerated Model for Multivariate Forecasting
Technical requirements
From univariate to multivariate – extending the prediction problem
Building and training the multivariate neural architecture
Enabling GPU execution for neural networks
Setting up a new GPU Python environment
Switching Python environments dynamically
Building the deployment application
Summary
Questions
Chapter 14: Combining KNIME and H2O to Predict Stock Prices
Technical requirements
Introducing the stock price prediction problem
Describing the KNIME H2O Machine Learning Integration
Starting a workflow running on the H2O platform
Introducing the H2O nodes for machine learning
Accessing and preparing data within KNIME
Accessing stock market data from Yahoo Finance
Preparing the data for modeling on H2O
Training an H2O model from within KNIME
Optimizing the number of predictor columns
Training, applying, and testing the optimized model
Consuming the H2O model in the deployment application
Summary
Questions
Final note
Answers
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Chapter 9
Chapter 10
Chapter 11
Chapter 12
Chapter 13
Chapter 14
Index
About Packt
Other Books You May Enjoy