An experimental evaluation of neural networks for classification

doi:10.1016/0305-0548(93)90063-O

Computers & Operations Research

Volume 20, Issue 7, September 1993, Pages 769-782

https://doi.org/10.1016/0305-0548(93)90063-O Get rights and content

Abstract

Artificial neural networks are new methods for classification. In this paper, we describe how to build neural network models. These models are then compared with classical models such as linear discriminant analysis and quadratic discriminant analysis. While neural network models can solve some difficult classification problems where classical models cannot, the results show that even under best conditions for the classical models, neural networks are quite competitive. Furthermore, neural networks are more robust in that they are less sensitive to changes in sample size, number of groups, number of variables, proportions of group memberships, and degrees of overlap among groups.

References (43)

W.V. Gehrlein
General mathematical programming formulations for the statistical classification problem
Ops Res. Lett.
(1986)
J.W. Denton et al.
A neural network approach to the classification problem
Expert Syst. Applic.
(1990)
A.D. Gordon
Classification
(1981)
N. Capon
Credit scoring systems: a critical analysis
J. Market.
(1982)
S. Sharma et al.
Early warning indicators of business failures
J. Market.
(1980)
R.A. Walking
Predicting tender offer success: a logistic analysis
J. Finance Quant. Analysis
(1985)
R.Y. Awh et al.
A discriminant analysis of economic, demographic, and attitudinal characteristics of bank charge-card holders: a case study
J. Finance
(1983)
P.O. Duda et al.
Pattern Classification and Scene Analysis
(1973)
R.A. Fisher
The statistical utilization of multiple measurements
Annals Eugen.
(1938)
C.A.B. Smith
Some examples of discrimination
Annals Eugen.
(1946)

R. Lippmann

An introduction to computing with neural nets

IEEE ASSP Mag.

(1987)

J.C. Hoskins et al.

Incipient fault detection and diagnosis using artificial neural networks

K.I. Lang et al.

Learning to tell two spirals apart

Y. Le Cun et al.

Handwritten digit recognition with a back-propagation network

Adv. Neural Inform. Process. Syst.

(1990)

G.L. Martin et al.

Recognizing hand-printed letters and digits

Adv. Neural Inform. Process. Syst.

(1990)

M.T. Leung et al.

Fingerprint processing using backpropagation neural networks

D.B. Malkoff

A neural network for real-time signal processing

Adv. Neural Inform. Process. Syst.

(1990)

T.J. Sejnowski et al.

Combining visual and acoustic speech signals with a neural network improves intelligibility

P.M. Shea et al.

Operational experience with a neural network in the detection of explosives in checked airline baggage

J.C. Singleton

Neural nets for bond rating improved by multiple hidden layers

K.Y. Tam et al.

Managerial applications of neural networks: the case of bank failure predictions

Mgnt Sci.

(1992)

Cited by (56)

MLR and ANN models of significant wave height on the west coast of India
2012, Computers and Geosciences
Citation Excerpt :
They pointed out that ANN can provide a good alternative to statistical regression, time series analysis and numerical methods. Comparison studies between statistical methods and ANN have been carried out by many authors (Aslanargun et al., 2007; Kumar, 2005; Shuhui, Wunsch, Hair, and Giesselmann, 2001; Warner and Misra, 1996; Subramanian, Hung, and Hu, 1993; Patuwo, Hu, and Hung, 1993; Hruschka, 1993; Tam and Kiang, 1992; Salchenberger, Cinar, and Lash, 1992; Tam, 1991). Mahjoobi and Etemad-Shahidi, 2008, 2009 investigated the performances of classification and regression trees.
Multiple linear regression (MLR) and artificial neural network (ANN) models are used in the present work to describe the significant wave height off Goa, located in the west Indian coast. A comparison study was carried out with the purpose of verifying when the artificial neural network and multiple linear regression models are appropriate for prediction of the significant wave height. Discussions of advantages and disadvantages are given in different point of view for both the methods. Several meteorological factors are used during the analysis and the ones affecting more to the model are kept. We concluded that non-linear models with wind speed and wind gust at a previous time step and air pressure, water temperature and air temperature at the same time step yield to better significant wave height models.
A tuning method for the architecture of neural network models incorporating GAM and GA as applied to bankruptcy prediction
2012, Expert Systems with Applications
Citation Excerpt :
For example, Salchenberger et al. (1992) recommend that the number of hidden nodes should be 75% of the number of input variables. Subramanian, Ming, and Hu (1993) indicated that the number of nodes in a single hidden layer should range from the number of output nodes to the number of input variables plus one. Later, more systematic methods were suggested by Masters (1993) and Torsun (1996), who proposed a geometric progression rule.
The performance of a neural network model is affected by important constituent elements such as input variables, the number of hidden nodes, and the value of the decay constant. This paper suggests a new approach to fine-tune these factors to improve their accuracy. For the input variable selection, the generalized additive model (GAM) is applied. The grid search method and the genetic algorithm are sequentially implemented to fine-tune the number of hidden nodes and the value of the weight decay parameters. This suggested method to improve the neural network model is used to predict the probability that a firm may apply for bankruptcy, and its performance is compared with the results of existing bankruptcy forecasting models such as case-based reasoning, the decision tree, the GAM, the generalized linear model, the multi-variate discriminant analysis, and the support vector machine. Our empirical results indicate that the newly tuned neural network model significantly outperforms the other models.
Comparison of neural networks and regression analysis: A new insight
2005, Expert Systems with Applications
Using neural networks for identifying organizational improvement strategies
2002, European Journal of Operational Research
The struggle to remain competitive in the global marketplace occupies much of the energy of today's firms. Although there are many performance improvement strategies that can be implemented by an organization, it is yet to be determined which strategy or combination of these strategies, if any, is most helpful in improving performance. A genetic algorithm trained neural network is used to identify such combinations to provide direction to managers as to which performance improvement strategies are associated with increases in performance as reported by operations managers. This study shows that this neural network approach can identify combinations that result in better approximations of performance, as compared to standard statistical techniques, and should be considered as an appropriate tool for performance improvement strategy selection.
A computational study on the performance of artificial neural networks under changing structural design and data distribution
2002, European Journal of Operational Research
Citation Excerpt :
Our conjecture is open to future tests, however. The number of classification groups, together with the number of attributes and prediction variables, indicate a measure of task complexity [35]. The task complexity is a consideration in the choice of a discriminant analysis technique.
This paper provides the results of our computational studies on artificial neural networks (ANNs) under various structural design and data distributions. A two-group classification problem is investigated. Simulated data with varying kurtosis and variance are used to examine how the ANN performs with respect to certain structural design (size and addition of input and weight noise) characteristics. The results of our study indicate that additive noise, size, and data distribution characteristics play an important role in learning, reliability and predictive ability of ANNs.
Determination of composition of mixed biological samples using laser-induced fluorescence and combined classification/regression models
2021, European Physical Journal Plus

View all citing articles on Scopus

^†: V. Subramanian received a Ph.D. in Business from Kent State University. His research interests include neural network theory and applications, object oriented systems, database theory and applications of AI to business.

^§: His research interests include applications of neural networks, applied statistics, marketing research and international business.

View full text

An experimental evaluation of neural networks for classification

Abstract

Ops Res. Lett.

Expert Syst. Applic.

Classification

Credit scoring systems: a critical analysis

J. Market.

Early warning indicators of business failures

J. Market.

Predicting tender offer success: a logistic analysis

J. Finance Quant. Analysis

A discriminant analysis of economic, demographic, and attitudinal characteristics of bank charge-card holders: a case study

J. Finance

Pattern Classification and Scene Analysis

The statistical utilization of multiple measurements

Annals Eugen.

Some examples of discrimination

Annals Eugen.

An introduction to computing with neural nets

IEEE ASSP Mag.

Incipient fault detection and diagnosis using artificial neural networks

Learning to tell two spirals apart

Handwritten digit recognition with a back-propagation network

Adv. Neural Inform. Process. Syst.

Recognizing hand-printed letters and digits

Adv. Neural Inform. Process. Syst.

Fingerprint processing using backpropagation neural networks

A neural network for real-time signal processing

Adv. Neural Inform. Process. Syst.

Combining visual and acoustic speech signals with a neural network improves intelligibility

Operational experience with a neural network in the detection of explosives in checked airline baggage

Neural nets for bond rating improved by multiple hidden layers

Managerial applications of neural networks: the case of bank failure predictions

Mgnt Sci.