|
view
back cover |
|
|
A quick peek
of the book
·
Table
of Contents
·
Introduction
·
Sample
Chapter 3.3
·
Sample
Chapter 3.17 |
|
Contents
of my Book to this Page |
|
|
|
SAS
Neural
Network Links |
|
·
SAS Neural Network Links |
|
|
|
Highlights
of the Book |
|
· |
An
overview to traditional regression modeling with an explanation of
the various assumptions that must be satisfied in both traditional
regression modeling and neural network modeling with a brief summary
of the various diagnostic statistics that are used to either detect
extreme values or influential data points along with the various
goodness-of-fit statistics that are used in determining the accuracy
of the modeling fit.
|
|
· |
An overview to neural network
modeling. The book will introduce the readers to the numerous
configuration settings of the various layers in neural network
modeling and the MLP architecture. The book will then briefly
explain the various neural network architectural designs and the
different combination, activation and error functions that can be
applied in Enterprise Miner and the nonlinear modeling designs. The
following sections will explain the hold out procedure that is
typically performed in generating unbiased modeling estimates and
the back-propagation iterative procedure that is used in calculating
the weight estimates in neural network modeling.
|
|
· |
An
overview to the various optimization techniques along with numerical
examples of the some of the optimization techniques that are used in
Enterprise Miner and neural network modeling.
|
|
· |
An overview to the various techniques
in measuring the importance of the input or predictor variables to
the neural network model. This will be followed by presenting the
readers to various ideas in developing a well-designed neural
network model and the advantages and disadvantages in neural network
modeling.
|
|
· |
An overview
to the SAS neural network modeling procedure called the NEURAL
procedure.
|
|
· |
An
overview to the powerful SAS product called Enterprise Miner.
|
|
· |
Step-by-step
instructions in constructing a Enterprise Miner process flow diagram
in order to perform neural network predictive modeling and
traditional regression modeling.
|
|
· |
An overview to the configuration
settings and listed results from the various nodes in Enterprise
Miner that are used in neural network modeling and traditional
regression modeling.
|
|
·
|
Comparing
neural network predictive modeling estimates with the more popular
statistical modeling estimates such as multiple linear regression
modeling, logistic regression modeling, non-linear regression
modeling, time series modeling and discriminant analysis based on
various examples from SAS manuals and literature with an added
overview to the corresponding statistical modeling designs and a
brief explanation to the associated SAS modeling procedures, option
statements and procedure output listings.
|
|
|
Overview
of the Book |
|
“Neural
Network Modeling using SAS Enterprise Miner” introduces the readers to a
non-linear modeling design called neural network modeling using SAS
Enterprise Miner v4.3. This book will reveal the power and ease of use of
the powerful new module in SAS with step-by-step instructions in creating
a process flow diagram in preparation to data-mining analysis and neural
network statistical modeling.
Enterprise
Miner is specifically designed for data mining that has been available
since SAS version 8 was released. Data mining is an analytical tool that
is used in solving critical business decisions by analyzing large amounts
of data that is designed to discover relationships and unknown patterns in
the data. Enterprise Miner is built around the data mining SEMMA
methodology that is
specifically designed in handling enormous data sets in preparation to
subsequent data analysis. In SAS Enterprise Miner, the abbreviation SEMMA
stands for Sampling, Exploring, Modifying, Modeling and Assessing large
amounts of data. Neural network modeling with regard to the data mining
tasks falls under nonlinear regression modeling or classification
modeling. That is, neural network modeling can perform both predictive
modeling and classification modeling that depends on the variable that we
want to predict either being continuous or categorical.
There
are several advantages to using neural network
modeling. That is, it’s
extremely flexibility in modeling a wide variety of statistical models and
interpolating complex nonlinear functions. One of the advantages in neural network modeling is that it does not require any distributional
assumptions between the response variable
and
the predictor variables to the model.
That is, unlike traditional regression modeling, neural network modeling
does not require the use of a previously selected mathematical functional
form between the variable that we want to predict and the predictor
variables to the model. And yet, neural network modeling is extremely
accurate in interpolating a wide variety of functional forms assuming that
there are a sufficient number of hidden layer units, an adequate amount of
data and a reasonable amount of computational time. In classification
modeling, neural network modeling can approximate any decision boundary by assigning the given data
into distinct groups with great precision.
SAS
Enterprise Miner is a very easy to learn and very easy to use. Users do
not even need to know SAS programming and have very little statistical
expertise in designing an Enterprise Miner project in order to develop a
completely comprehensive statistical analysis reporting system. Yet, an
expert statistician can adjust the default settings and run the Enterprise
Miner process flow diagram
to their own personal
specifications. It takes advantage of the intuitive point-and-click
programming within a convenient graphic user interface. The diagram
workspace or the process flow diagram has the feel and same appearance
much like the desktop environment in Microsoft Windows. Users have various
icons to their disposal that will perform a wide variety of statistical
analysis. The nodes can be simply dragged onto the diagram workspace that
are then connected to one another all within the same graphical diagram
workspace.
This
book consists of step-by-step instructions along with an assortment of
illustrations for the reader to get familiar with the various nodes and
the corresponding working environment in SAS Enterprise Miner. There are numerous examples in explaining the various complex neural
network designs and optimization techniques used in neural network
modeling with numerous examples taken from various SAS literature
comparing the prediction estimates between both neural network and
traditional statistical modeling with an explanation to the listed results
that are generated between the models.
In
the introductory sections, the book will provide a brief description of
traditional regression modeling and the various assumptions that must be
satisfied. The following section of the book offers a graphical
illustration to the similarities between the parameter estimates in
multiple linear regression models and the weight estimates in neural
network models. Subsequent sections to the book introduces the reader to
the neural network design and the various
configuration settings such as the various neural network
layers, weight estimates, combination functions, transfer functions, objective or error functions and the optimization
techniques
that are used in neural
network modeling. “Neural Network Modeling using SAS Enterprise Miner”
will explain the various optimization techniques that are needed in
solving the non-linear neural network modeling design. Optimization
techniques such as the various line search
and grid search techniques
that are used within the neural network node in Enterprise Miner. It will
be followed by various numerical examples in order to simplify the
complexity of the various optimization techniques that are applied in
calculating the neural network weight estimates that are designed in
finding the smallest error to the neural network model. The book will then
explain the various ways in constructing a well-designed neural network
model and conclude the section in explaining to the readers of the
advantages and disadvantages to neural network modeling.
The
following section will then introduce the readers to SAS Enterprise Miner.
The section will begin with a brief overview to the SAS Enterprise Miner
interface that will explain the main menu option settings and the various
start-up configuration settings to the currently opened Enterprise Miner
project. The section will explain the basic steps in constructing
Enterprise Miner project and the corresponding process flow diagrams. The
book then leads into a general overview of the purpose of the available
nodes in Enterprise Miner for data mining analysis with regard to the
SEMMA design.
The
following chapter will explain the configuration settings and the
corresponding results that are generated from the Input Data Source node,
Data Partition node, Regression node, Neural Network node,
SAS Code node,
Assessment node and the Reporter node. These nodes are used to construct
the process flow diagram to compare neural network estimates with the
traditional statistical modeling designs. Each section of the book will
explain the purpose of the numerous tabs within each node along with the
various option settings that are listed within each tab of the
corresponding the node followed by an explanation of the results that are
generated from each one of the nodes.
The
final chapter to the book will compare the neural network estimates with the various
traditional statistical modeling designs such as linear and non-linear
regression modeling, logistic regression modeling, time series modeling
and discriminant analysis. Each section provides a brief summary of the
various statistical modeling designs and the associated SAS procedure and
option statements that are needed in producing the corresponding modeling
results. The various statistical modeling comparison examples that are
presented are from various SAS manuals and literature that will enable the
reader to cross-reference the statistical results. The book will compare
the accuracy in the fit between both models by graphically comparing the
modeling estimates and analyzing the various modeling assessment
statistics
from both models in
determining the best fit.
|
|
|
|
New Articles |
|
Leon
L. Fedenczuk,To
Neural or Not to Neural
This
paper compares solutions from neural
networks to solutions from logistic
regression, discriminant analysis, and decision tree analysis
for identifying the most profitable customers.
|
|
Lee,
Carl, Rey, Tim, Mentele, James, Garver, James, Neural Network
Techniques for Modeling Loyalty and Profitabilty
This
article presents a new technique namely structured Neural Network (SNN)
technique for modeling loyalty and profitability and demonstrates an
application for a chemical company.
|
|
McNelis,
Paul, Nickelsburg, Jerry, Neural
Networks and Genetic Algorithms as Tools for Forecasting Demand in
Consumer Durables
This
paper explains the power of neural network
modeling is predicting consumer demands for automobiles.
|
|
|
|
Articles of Interest
|
|
Sarle,
Warren, Neural
Networks and Statistical Models
This
is one of the best articles in comparing the similarities between
traditional statistical modeling and neural network modeling. If you are
familiar with the various traditional statistical modeling techniques,
then the article will enable you to understand neural network modeling
design much easier.
|
|
Sarle,
Warren, Neural
Network FAQ
This
article is one of the best articles available for neural
network modeling. The article provides various questions in regards to
neural network modeling.
|
|
Sarle,
Warren, Neural
Network Implementation in SAS Software
This
article provides information on the various optimization techniques that
are used in neural network modeling and SAS®
Enterprise Miner.
|
|
Sarle,
Warren, Kangaroos
and Training Neural Networks
This
article by Warren Sarle explains the various minimization techniques that
are used in neural network modeling in finding the minimum error by making
the analogy of a kangaroo hopping around a mountain.
|
|
Sarle,
Warren, Prediction
with Missing Inputs
The
purpose of the paper is briefly explain the various imputation techniques
with missing inputs in the prediction model.
|
|
Sarle,
Warren, Stopped
Training and Other Remedies for Overfitting
The
paper is to provide various ways in determining appropriate times to stop
network training and ways to handling overfitting in the neural network
model based on various simulated models by comparing the number of hidden
units, the generalization error, and various modeling selection
techniques.
|
|
Sarle,
Warren, Stopped
Training and Other Remedies for Overfitting
The
paper is to provide various ways in determining appropriate times to stop
network training and ways to handling overfitting in the neural network
model based on vari
|
|
Sarle,
Warren, Using
IML with Feed-Forward Neural Networks
This
short paper provides you with
SAS IML code to construct MLP neural network models.
|
|
Sarle,
Warren, Neural
Networks vs. Statistics
This
short paper provides a brief
summary of explaining various methods of estimating generalization error.
|
|
Bishop, Christopher, Pattern
Recognition and Feed-forward Networks
This
article by Christopher Bishop briefly explains the feed-forward neural
network model and its architectural design.
|
|
Bishop, Christopher, Neural Networks
This
article by Christopher Bishop provides an overview to various aspects to
neural network modeling from the architectural design, objective
functions, cost functions, optimization algorithms, and regularization.
|
|
Bishop, Christopher, Exact
Calculation of Hessian Matrix for Multi-Layer Perceptron
This
article by Christopher Bishop briefly introduces the back-propogation
algorithm that is based on the elements from the Hessian matrix. The
Hessian matrix consists of first and second derivatives of the error
function that is used in determining the weight estimates at each
iteration from network training.
|
|
Sarma, Kattamuri S , Using
SAS Enterprise Miner for Forecasting
This
article explains the process of constructing a process flow diagram to
construct neural network time series forecasting models using SAS®
Enterprise Miner.
|
|
University
of Alabama, Introduction to
SAS Enterprise Miner
The
paper sets-up a workflow diagram to generate various plots in the
MultiPlot node.
|
|
Commerce
Data Mining Techniques
The
paper exposes the readers to the working environment of SAS®
Enterprise Miner with a brief explanation of the various nodes and the
corresponding tabs.
|
|
Wahab, Diana, Neural
Network Models Using SAS Enterprise Miner
This
article provides information in regards to the advantages to neural
network modeling used in fitting time series data.
|
|
Neural
Network Modeling using SAS Enterprise Miner Blog
|
|
Neural
Network Modeling using SAS Enterprise Miner Blog
The
following link is to my personal blog where you may post any questions that you
might have with my book, that is "Neural Network Modeling using SAS®
Enterprise Miner". Otherwise, please feel free to ask any questions that
you might have with SAS® Enterprise Miner.
|
|
|
|
Ordering
Information
|
|
authorHOUSE
Sales
Department
1663
Liberty Drive, Suite 200
Bloomington,
IN 47403
To
order call: (888) 280-7715, visit http://www.authorhouse.com
or e-mail: bkorders@authorhouse.com
|
|
|
|
|
|
Back
to Home Page
|
