Since the book's appearance in early 2007, it has been used in many classes, ranging from dedicated data mining classes to more general business intelligence courses. Following feedback from instructors teaching both MBA and undergraduate courses, as well as students, we revised some of the existing chapters as well as covered two new topics that are central in data mining: data visualization and time series forecasting.
We have added a set of three chapters on time series forecasting (Chapters 15–17), which present the most commonly used forecasting tools in the business world. They include a set of new datasets and exercises, and a new case (in Chapter 18).
The chapter on data visualization provides comprehensive coverage of basic and advanced visualization techniques that support the exploratory step of data mining. We also provide a discussion of interactive visualization principles and tools, and the chapter exercises include assignments to familiarize readers with interactive visualization in practice.
In the new edition we have created separate chapters for the k-nearest-neighbor and naive Bayes methods. The explanation of the naive Bayes classifier is now clearer, and additional exercises have been added to both chapters.
Another addition are brief chapter summaries at the beginning of each chapter.
We have also reorganized the order of some chapters, following readers' feedback. The chapters are now grouped into seven parts: Preliminaries, Data Exploration and Dimension Reduction, Performance Evaluation, Prediction and Classification Methods, Mining Relationships Among Records, Forecasting Time Series, and Cases. The new organization is aimed at helping instructors of various types of courses to choose subsets of topics to teach.
- Two-semester data mining courses
could cover in detail data exploration and dimension reduction and supervised learning in one term (choosing the type and amount of prediction and classification methods according to the course flavor and the audience interest). Forecasting time series and unsupervised learning can be covered in the second term.
- Single-semester data mining courses
would do best to concentrate on the first parts of the book, and only introduce time series forecasting as time allows. This is especially true if a dedicated forecasting course is offered in the program.
- General business intelligence courses
would best focus on the first three parts, then choose a small number of prediction/classification methods for illustration, and present the mining relationships chapters. All these can be covered via a few cases, where students read the relevant chapters that support the analysis done in the case.
- A set of data mining courses
that constitute a concentration can be built according to the sequence of parts in the book. The first three parts (Preliminaries, Data Exploration and Dimension Reduction, and Performance Evaluation) should serve as requirements for the next courses. Cases can be used either within appropriate topic courses or as project-type courses.
In all courses, we strongly recommend including a project component, where data are either collected by students according to their interest or provided by the instructor (e.g., from the many data mining competition datasets available). From our experience and other instructors' experience, such projects enhance the learning and provide students with an excellent opportunity to understand the strengths of data mining and the challenges that arise in the process.