What this book covers
Chapter 1, Introducing Time Series Analysis, explains what a time series is, states some classic time series problems, and introduces the two historical approaches: statistics and machine learning.
Chapter 2, Introduction to KNIME Analytics Platform, explains the basic concepts of KNIME Analytics Platform and its time series integration. This chapter covers installation, an introduction to the platform, and a first workflow example.
Chapter 3, Preparing Data for Time Series Analysis, introduces the common first steps in a time series analysis project. It explores different sources of time series data and shows time alignment, time aggregation, and missing value imputation as common preprocessing steps.
Chapter 4, Time Series Visualization, explores time series visualization. It provides an exploration of the most common visualization techniques to visually represent and display the time series data: from the classic line plot to the lag plot, and from the seasonal plot to the box plot.
Chapter 5, Time Series Components and Statistical Properties, introduces common concepts and measures for descriptive statistics of time series, including the decomposition of a time series, autocorrelation measures and plots, and the stationarity property.
Chapter 6, Humidity Forecasting with Classical Methods, completes a classic time series analysis use case: forecasting. It introduces some simple yet powerful classical methods, which often solve the time series analysis problem quickly without much computational expense.
Chapter 7, Forecasting the Temperature with ARIMA and SARIMA Models, delves into the ARIMA and SARIMA models. It aims at predicting tomorrow’s temperatures with the whole range of ARIMA models: AR, ARMA, ARIMA, and SARIMA.
Chapter 8, Audio Signal Classification with an FFT and a Gradient Boosted Forest, introduces a use case for signal classification. It performs the classification of audio signals via a Gradient Boosted Forest model and the FFT transforms the raw audio signals before modeling.
Chapter 9, Training and Deploying a Neural Network to Predict Glucose Levels, gives an example of a critical prediction problem: predicting the glucose level for a timely insulin intervention. This chapter also introduces neural networks.
Chapter 10, Predicting Energy Demand with an LSTM Model, introduces recurrent neural networks based on Long Short Term Memory (LSTM) layers, which are advanced predictors when temporal context is involved. It tests whether the prediction accuracy improves considerably from an ARIMA model when using a recurrent LSTM-based neural network.
Chapter 11, Anomaly Detection – Predicting Failure with No Failure Examples, tackles the problem of anomaly detection in predictive maintenance by introducing approaches that work exclusively on the data from a correctly working system.
Chapter 12, Predicting Taxi Demand on the Spark Platform, implements a solution to the demand prediction problem via a Random Forest to run on a Spark platform in an attempt to make the solution more scalable.
Chapter 13, GPU Accelerated Model for Multivariate Forecasting, extends the demand prediction problem to a multivariate by taking into account exogenous time series as well, and scalable, by training the recurrent neural network on a GPU-enabled machine.
Chapter 14, Combining KNIME and H2O to Predict Stock Prices, describes the integration of KNIME Analytics Platform with H2O, another open source platform, to implement a solution for stock price prediction.