Pm. West et al., A COMPARATIVE-ANALYSIS OF NEURAL NETWORKS AND STATISTICAL-METHODS FORPREDICTING CONSUMER CHOICE, Marketing science, 16(4), 1997, pp. 370-391
This paper presents a definitive description of neural network methodo
logy and provides an evaluation of its advantages and disadvantages re
lative to statistical procedures. The development of this rich class o
f models was inspired by the neural architecture of the human brain. T
hese models mathematically emulate the neurophysical structure and dec
ision making of the human brain, and, from a statistical perspective,
are closely related to generalized linear models. Artificial neural ne
tworks are, however, nonlinear and use a different estimation procedur
e (feed forward and back propagation) than is used in traditional stat
istical models (least squares or maximum likelihood). Additionally, ne
ural network models do not require the same restrictive assumptions ab
out the relationship between the independent variables and dependent v
ariable(s). Consequently, these models have already been very successf
ully applied in many diverse disciplines, including biology, psycholog
y, statistics, mathematics, business, insurance, and computer science.
We propose that neural networks will prove to be a valuable tool for
marketers concerned with predicting consumer choice. We will demonstra
te that neural networks provide superior predictions regarding consume
r decision processes. In the context of modeling consumer judgment and
decision making, for example, neural network models can offer signifi
cant improvement over traditional statistical methods because of their
ability to capture nonlinear relationships associated with the use of
noncompensatory decision rules. Our analysis reveals that neural netw
orks have great potential for improving model predictions in nonlinear
decision contexts without sacrificing performance in linear decision
contexts. This paper provides a detailed introduction to neural networ
ks that is understandable to both the academic researcher and the prac
titioner. This exposition is intended to provide both the intuition an
d the rigorous mathematical models needed for successful applications.
in particular, a step-by-step outline of: how to use the models is pr
ovided along with a discussion of the strengths and weaknesses of the
model. We also address the robustness of the neural network models and
discuss how far wrong you might go using neural network models versus
traditional statistical methods. Herein, we report the results of two
studies. The first is a numerical simulation comparing the ability of
neural networks with discriminant analysis and logistic regression at
predicting choices made by decision rules that vary in complexity. Th
is includes simulations involving two noncompensatory decision rules a
nd one compensatory decision rule that involves attribute thresholds.
In particular, we test a variant of the satisficing rule used by Johns
on et al. (1989) that sets a lower bound threshold on all attribute va
lues and a ''latitude of acceptance'' model that sets both a lower thr
eshold and an upper threshold on attribute values, mimicking an ''idea
l point'' model (Coombs and Avrunin 1977). We also test a compensatory
rule that equally weights attributes and judges the acceptability of
an alternative based on the sum of its attribute values. Thus, the sim
ulations include both a linear environment, in which traditional stati
stical models might be deemed appropriate, as well as a nonlinear envi
ronment where statistical models might not be appropriate. The complex
ity of the decision rules was varied to test for any potential degrada
tion in model performance. For these simulated data it is shown that,
in general, the neural network model outperforms the commonly used sta
tistical procedures in terms of explained variance and out-of-sample p
redictive accuracy. An empirical study bridging the behavioral and sta
tistical lines of research was also conducted. Here we examine the pre
dictive relationship between retail store image variables and consumer
patronage behavior. A direct comparison between a neural network mode
l and the more commonly encountered techniques oi discriminant analysi
s and factor analysis followed by logistic regression is presented. Ag
ain the results reveal that the neural network model outperformed the
statistical procedures in terms of explained variance and out-of-sampl
e predictive accuracy. We conclude that neural network models offer su
perior predictive capabilities over traditional statistical methods in
predicting consumer choice in nonlinear and linear settings.