Introduction to Data Mining by Pang-ning Tan, Michael Steinbach, Vipin Kumar (2005) Paperback

If you think you are interested in Data Mining this is a great place to start. This book would work well for people interested in self study, or someone who is considering going to grad school to pursue a field utilizing data mining, or doing data mining research directly. The book covers the core data mining concepts, with clear examples on how the concepts could be applied to toy problems. The book is light on math and heavy on application, which is great at maintaining interest. This book is not commonly used as a course textbook at the grad level because of its shallow treatment of the underlying math. Sometimes you just want to know how, and worry about why later. Also, if you think data mining might be of use to your research or your professional work, this book provides a broad overview of topics. If you are unfamiliar with data mining, and have just heard the term, running through the introductions of each chapter will quickly point you to techniques that will be most useful to you.

This book is amazingly well written. Everything is explained in a very clear and to-the-point style. The book can be read from front to back or used as a reference book. It contains countless diagrams and the structure of the content is immediately apparent. The book covers a lot of the important aspects of data mining. It provides algorithms and techniques for classification, clustering, association analysis, and anomaly detection. Every algorithm is not only formally stated, but also explained in a way that conveys intuition. I only wish other authors also wrote books this way.

I've just made it through the first 6 chapters of the book so far but I really enjoy this book so far. This book is terrific at introducing this material in an easy-to-understand manner. I've found myself using to supplement my machine learning textbook when more thorough explanations are needed. The section on support vectors was the easiest to grasp from about a dozen references I had on hand. I've seen a few typos here and there but I suppose that's expected from a first edition.

I decided to start with this book as I think it is the most convenient to start in the data mining field. One big advantage of the book is the way data mining techniques are explained. It is mainly based on textual and graphical explanations. There is little equations, only what is necessary to implement the algorithms. This book widely cover areas such as data preparation and understanding, classification, anomaly detection, association analysis and clusering. Although the book has a strong emphasis on the two last ones, nearly all standard data mining techniques are at least briefly discussed. However, this book does only have a fiew pages about kernel methods for example. Indeed, it is normal, as kernel methods are more suitable for machine learning (I mean making prediction) than data mining (I mean looking for description). Therefore, this book is: * able to explain data mining without thousands of equations * a good way to start with data mining * covering nearly all standard data mining techniques * focused on association analysis and clustering and it is not: * a good book for kernel methods and other advanced techniques * written in the statistical nor in the database perspective My comment: if you are in the data mining field and not comming from mathematics or databases, then you really should buy this book.

As databases keep growing unabatedly, so too has the need for smart data mining. For a competitive edge in business, it helps to be able to analyse your data in unique ways. This text gives you a thorough education in state of the art data mining. Appropriate for both a student and a professional in the field. The extensive problem sets are well suited for the student. These often expand on concepts in the narrative, and are worth tackling. The central theme in the book is how to classify data, or find associations or clusters within it. Cluster analysis gets two chapters that are superbly done. These summarise decades of research into methods of grouping data into clusters. Usually hard to do, because an element of subjectivity can creep into the results. If your data is scattered in some n-dimensional space, then clusters might exist. But how to find them? The chapters show that the number of clusters and the constituents of these can depend on which method you adopt, and various initial conditions, like [essentially] seed values for clusters, if you choose a prototype cluster method like K-means. The descriptions of the cluster algorithms are succinct. Why is this useful? Because it helps you easily understand the operations of the algorithms, without drowning you in low level detail. Plus, by presenting a meta-level comparison between the algorithms, you can develop insight into rolling your own methods, specific to your data. Part of my research involves finding new ways to make clusters, and the text was very useful in explaining the existing ideas.