Восстановить пароль
FAQ по входу

Taguchi Genichi, Jugulum Rajesh. The Mahalanobis-Taguchi Strategy: A Pattern Technology System

  • Файл формата pdf
  • размером 7,90 МБ
  • Добавлен пользователем
  • Отредактирован
Taguchi Genichi, Jugulum Rajesh. The Mahalanobis-Taguchi Strategy: A Pattern Technology System
Wiley, 2002. — 256 p. — ISBN-10 0471023337; ISBN-13: 978-0471023333.
Cutting-edge measurement technology for multidimensional systems
The Mahalanobis-Taguchi Strategy presents methods for developing multidimensional measurement scales that are up to date with the most current trends in multivariate diagnosis/pattern recognition-namely, using measures and procedures that are data analytic and not dependent upon the distribution of the characteristics defining the system. Applications for these measurement scales are also explored across a wide range of disciplines from manufacturing to medicine.
This book presents methods that integrate mathematical and statistical concepts such as Mahalanobis distance and Gram-Schmidt's orthogonalization method with the principles of Taguchi methods. These completely new systems of measurement and analysis move beyond anything Dr. Taguchi has done in the past. Coverage includes the refined Mahalanobis-Taguchi system, the Mahalanobis-Taguchi-Gram-Schmidt method, the Adjoint Matrix method, and other advanced topics, along with a detailed examination of each method. In addition to examining how real-world problems are solved using these methods, critical comparisons are made between the methods covered here and existing multivariate diagnosis/pattern recognition techniques.
The Mahalanobis-Taguchi Strategy: A Pattern Technology System is an essential book for engineers, designers, and statistical quality experts and programmers in the fields of engineering and computer science, as well as researchers in finance, medicine, statistics, and general science.
This book (The Mahalanobis-Taguchi Strategy) describes the best multi-dimensional database system that exists today. This system takes a database consisting of many variables used to describe a certain `reference population', and converts it into just one variable called MD (the Mahalanobis Distance). The value of MD tells us whether a particular member of a population either belongs to the reference population, or whether that particular member is a member of some other population. Both `populations' are defined by the user.
[A Note: I used this strategy on a large group of buyers who were `loyal' to my client's international product - and also on a large group of buyers of the same type of product who were `disloyal' to my client's international product (meaning that they regularly purchased a competitor's brand). The Mahalanobis-Taguchi Strategy reliably and effectively sorted members of the two groups. Furthermore, MTS (as I refer to it) was capable of identifying several dozen variables which were the keys to separating the two groups. As a matter of fact, the MD values for the disloyal group members generally disclosed the `degree' of disloyalty. Another benefit: the results of the MTS experiment were used to re-vamp and develop an entirely new (and far more effective) marketing strategy for this multi-national corporation.]
A reader of this book will soon learn that it is really an `advanced' treatment of the subject - not necessarily meant for an MTS novice. For example, besides MTS, the authors introduce an alternative method which they call MTGS (for "Mahalanobis-Taguchi Gram-Schmidt"), which they say `sometimes' is advantageous to employ, instead of MTS alone. Another value of this book is that it contains arguments primarily against using the older methods of multivariate analysis (or pattern recognition) - including methods like Principal Component Analysis, Discrimination and Classification Method, Multiple Regression Analysis, Stepwise Regression, Test of Additional Information (Rao's Test), Multivariate Control Charts and Artificial Neural Networks.
Alas! A beginner or novice can receive some benefits from reading this book - like hearing the overall description of MTS, and like moving through several international case studies. However, the book "really" does not tell the reader the secrets of `how' to apply MTS to a particular multi-variable database problem, or how to design an MTS experiment. Even the MTS experiments have trouble, for example, in designing appropriate questionnaires for an MTS marketing experiment - questions in such questionnaires demand that the developers have not only an excellent understanding of the product or service involved, but also a fairly deep understanding of human psychology, coupled with the ability to translated psychological characteristics into meaningful questions (I only know of one advertising agency in the world who can do this).
Based upon the contents of this book (and a companion book), I have very effectively applied MTS to several diverse areas, where pattern-recognition was the objective. These included the following:
* Marketing - Here we examined the market for a huge product used by mothers for young children. MTS was able to decisively discriminate between mothers who regularly purchased one brand, from mothers who purchased another brand. Furthermore, we identified several dozen chief characteristics of both groups of mothers. The results were invaluable for future marketing and advertising campaigns.
* Loyalty Programs - Here we were able to determine the key variables responsible for keeping current customers loyal, as well as the variables which were important to competitors' loyalty programs. MTS did this by considering not only the numerical values of the variables, but by considering their correlation structures as well.
* The Golf Swing - We applied MTS to the golf swing of excellent golfers and to beginning golfers, determining which biomechanical variables contributed to a golfer's ability to A) hit the ball straight towards the intended target, every time, B) maximize the range of each ball, without compromising down-range, side-to-side dispersion. This experiment took only two hours to conduct. We employed a system which measured biomechanics variables at every point in the swing, and we measured over 40 ball characteristics from impact all the way to the ball-landing point. The result was a vastly improved swing (more consistent, significantly reduced down-range side-to-side dispersion, and longer range).
* Equities Trading - We applied MTS to develop highly reliable buy and sell indicators. Five dozen variables were initially considered, but we were able to reduce the number of variables to a much smaller number. The system works well.
Overall, this system not only helped me to better-apply MTS to various pattern-recognition systems - it also served as an `idea kernel' for applying MTS to a host of other application areas. The book is definitely not for everyone, but it is a real gem.
PS: I would disagree with some of the comments on the back cover of this book: "The Mahalanobis--Taguchi Strategy: A Pattern Technology System is an essential book for engineers, designers, and statistical quality experts and programmers in the fields of engineering and computer science, as well as researchers in finance, medicine, statistics, and general science." Why? Because it is missing an all-important ingredient: the `How To', written in a simple way which engineers, designers and statistical quality experts and programmers can readily grasp. Perhaps another book will come along that does this.
  • Чтобы скачать этот файл зарегистрируйтесь и/или войдите на сайт используя форму сверху.
  • Регистрация