Book Review of Data Mining with Neural Networks

Data Mining with Neural Networks
Solving Business Problems -
from Application Development
to Decision Support

Joseph P. Bigus
McGraw-Hill
New York
1996

Reviewed by: Ronald D. Pollock
November 1996

TARGET AUDIENCE

Data Mining with Neural Networks targets executives, managers, and computer professionals with an explanation of data mining and neural networks from a business information and management prospective. The book focuses on the practical and competitive advantages provided by data mining and neural networks when used as strategic technologies within the business.

"The secret to business is to know something that nobody else knows." Aristotle Onassis

CONTENT

"Information networks straddle the world. Nothing remains concealed. But the sheer volume of information dissolves the information. We are unable to take it all in." Gunther Grass

According to Bigus, data mining is an efficient means of knowledge discovery if the information obtained is worth more than the cost of processing the raw data. Data mining, also referred to as knowledge discovery, is defined by Bigus as

" . . . the efficient discovery of valuable nonobvious information from a large collection of data" (p. 9).

Part 1: The Data Mining Process Using Neural Networks

Bigus establishes the framework for data mining as it has developed into a value added to the business activities of decision support and application development. He traces the advances in computer technology that resulted in the ability of organizations to store tremendous amounts of data online. This glut of data created a demand for data mining because traditional query and analysis methods could not handle the huge quantities of data.

The chapters in Part 1 take a step-by-step approach to developing an understanding of the methodologies of data mining. The chapters progress from a discussion of historical backgrounds through the deployment and maintenance of neural network applications. Part 1 ends with a discussion of the symbiotic relationship between intelligent agents and data mining: intelligent agents can control data mining, while data mining can add learning capabilities to intelligent agents.

Part 2: Data Mining Application Case Studies

Part 2 provides detailed examinations of data mining applications in business. The discussions focus on the business problem, the data, and the data mining process, not on the mechanics of using the tool to develop applications. The solutions use the neural network data mining methodology from Part 1. Bigus points out that, while the data mining applications discussed in the book use the IBM Neural Network Utility, the basic methodologies are similar regardless of the neural network product. The chapters on case studies include examples of:

Market segmentation: Segment or cluster the total market into specialized niches (target marketing).

Real estate pricing model: Determining the attributes that contribute to the market value of property.

Customer ranking model: Use current customer information to rank potential new customers.

Sales forecasting: Predict sales and inventory requirements to minimize inventory costs.

Additional features

The final portion of the book contains a glossary of terms used in neural network data mining and appendices with information about:

The IBM Neural Network Utility: The major features of the product and how they relate to the data minining methodology described in Part 1

Fuzzy Logic: An introduction to fuzzy sets, fuzzy logic, fuzzy rule systems, and the ways fuzzy logic and neural networks have been combined synergistically

Genetic Algorithms: An introduction to genetic algorithms and how they are used in neural networks

IMPRESSIONS

While Data Mining with Neural Networks targets the business audience, it also provides students with a comprehensive introduction to the background, methodologies, and applications of data mining with neural networks. The book is written in an informative, "how-to" format that is easy and enjoyable reading. The illustrations complement the text with relevant graphical representations. The case studies provide examples of "real world" types of information problems that neural network data mining can solve.

This is not a technical book. Readers will be disappointed if they expect to find source code and the inner workings of neural network data mining technology in Data Mining with Neural Networks. Also, those who are not fans of IBM may find objections with the book, for the author's association with IBM and with IBM's neural network data mining technology are evident. Bigus, however, provides much more than an IBM perspective, pointing out that the methodologies described in his book are not unique to IBM and are the basis for other neural network data mining tools.

The author states in his introduction that, upon completion of the book, the reader should "know what data mining is, what problems neural networks can solve today, how to determine if a problem is appropriate for a neural network solution, how to set up the problem for solution, and finally how to solve it." From this standpoint, the author achieves his intention with his target audience.

ADDITIONAL QUOTATIONS

Joseph P. Bigus included a number of great quotations in Data Mining with Neural Networks. In addition to the quotes at the beginning of this review, here are a few of those I particularly enjoyed.

"If you can look into the seeds of time and say which grain will grow, and which will not, speak then to me." Shakespeare, Mcbeth

"Man is still the most extraordinary computer of all." John F. Kennedy

"The real queston is not whether machines think, but whether men do." B.F. Skinner

"We want to replace the computer metaphor with the brain metaphor." David Rumelhart

"It is a capital mistake to theorize before one has data." Sir Arthur Conan Doyle

"A learning machine is any device whose actions are influenced by past experience." Nils Nilsson

"The future of computing will be 100% driven by delegating to, rather than manipulating, computers." Nicholas Negroponte

"Everything is worth what its purchaser will pay for it." Publilius Syrus

"When you've got them by their wallets, their hearts and minds will follow." Fern Naito

DATA MINING: ADDITIONAL RESOURCES

dbProphet White Paper. Trajecta.
White paper that explores artificial neural networks and their role as part of the adaptive technology architecture of the dbProphet predictive modeling tool.

Darwin. Thinking Machine Corporation.
A suite of software tools to analyze large databases to discover new patterns and predict future trends, using neural networks and other discovery methods.

Data Mining: An Introduction. The Queens University of Belfast.
General discussion of data mining, with neural networks described and illustrated in the section titled: Knowledge Representation Methods.

Neural Connection. SPSS, Inc.
State-of-the-art neural network power and flexibility for finding patterns in data and building better models.

Recon Data Mining System. Lockheed Martin Product and Services
A complete data mining solution employing both top-down and bottom-up data mining techniques.

Ultragem Data Mining (Genetic Algorithm Data Mining).
Descibes genetic algorithm data mining, including a FAQ and examples of business applications.

Direct comments to Ron Pollock, Doctoral Student, The University of Texas at Austin.

Go back to: Titles of Book Reviews | AI and Expert Systems | Tentative Calendar | GSLIS