Intelligent data analysis. An introduction

Authors

Publication date 1 September 1999

Bayesian method fuzzy logic theory Intelligent data analysis neural networks statistical inference time series

Collections of articles of miscellaneous specific interest (00B15) Bayesian inference (62F15) Research exposition (monographs, survey articles) pertaining to computer science (68-02) Proceedings, conferences, collections, etc. pertaining to computer science (68-06) Data structures (68P05)

Review text

We human beings love to collect data: grades recorded by universities, personal financial information maintained by credit agencies, mailing lists gathered by political campaigns, satellite images received by weather reporting agencies, etc. With the modern computers, it is relatively easy to collect all those data. The question is, of course, how to make an effective use of them, i.e., further analyze and summarize those data so that useful information can be derived. This book collects some of the essential data analytical techniques, both classic and modern, that the computer age data analysts can use to draw mathematically sound conclusions. After an introductory chapter, which discusses the nature of the data, as well as the impact of computer technology on data analysis, the main body of this book consists of eight chapters. Each of them covers a major data analytical technique, and is written by an individual who apparently knows his/her subject area inside out. Chapter 2 describes some basic concepts related to statistical inference such as data sampling, interval estimation, error predication, and various inference paradigms, and introduces some of the computationally intensive resampling methods, particularly, cross-validation and bootstrapping. Chapter 3 further discusses some of the other widely used statistical methods such as regression analysis, and those ``ought to be widely used ones such as various multivariate analysis, considering the role they might play in automatic data collection. Since all the corresponding algorithms are already implemented in various statistical packages, this chapter focuses on where and how to apply those methods and interpret the results. Chapter 4 turns to the increasingly popular Bayesian method, which incorporates some prior information about the collected data into the data analysis process to obtain updated posterior information about the data population. After introducing the basic concepts, and demonstrating this approach with some easier tasks such as data estimation, model selection, and reliability assessment, this chapter applies this method to more realistic situations such as multiple linear regression models and generalized linear models. It ends up with a discussion of the Bayesian Belief Networks and its application to data predication, classification, and modeling tasks. Chapter 5 studies various analytical techniques applicable to time series. After a brief review of the traditional linear model case and general theory on dynamic systems, this chapter discusses the delay-coordinate embedding technique, which allows one to construct a useful model of the internal dynamics of a nonlinear system, when only insufficient data specification exists. Chapter 6 then goes over the rule induction issue, i.e., based on a set of data sample, how to find simple rules that summarize all those examples and will, to the degree possible, correctly classify those yet unseen instances. Chapter 7 presents a detail overview of the well-known neural networks, which simulates the functions of biological organisms. It reviews some of the fundamental concepts, and discusses the most commonly used neural structure, i.e., the multilayer feed forward network, and the associated learning algorithms. Various other techniques are also studied, which helps to either discover any unknown relationship, or provide a better representation of some of the known ones, for the data population under study. This chapter also introduces some of the techniques that will generalize neural networks from a static structure to a dynamic one to enable us to study systems whose behaviors might be time dependent. Chapter 8 touches on another hot topic, namely, the fuzzy logic theory, which deals with inherently imprecise data. The main emphasis is on the generation of fuzzy rules, or equivalently, extraction of fuzzy graphs, or trees, from the data samples. To solve many of the optimization problems that often occur in the context of data analysis, Chapter 9 then presents some of the representative stochastic searching algorithms, mainly simulated annealing and evolutionary algorithms, and many of their cousins such as genetic algorithms and genetic programming. As an end note, Chapter 10 discusses a wide range of interesting data analysis applications of those covered techniques, and investigates some of the future development issues such as strategy, problem formulation, model construction, scalability, etc. This book concludes with a rich list of bibliography that contains approximately 450 titles, and a relatively simple index. It also contains a categorized collection of available data analytical packages, for each of which, the book gives a short description, its data analysis functionalities, platform(s), together with some contact information. Besides providing an up-to-date comprehensive review of some of the most important and widely used data analysis techniques for DA professionals, this clearly written and well edited book, with its many illustrations and simple, but telling, examples, will certainly be an excellent source book for those who are just entering the DA field, and particularly, for those who are taking a DA related course.

Recommendations

Cited in

(22)

This page was built for publication: Intelligent data analysis. An introduction

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4258213)