Two‐Group Classification Using Neural Networks*

Abstract

Artificial neural networks are new methods for classification. We investigate two important issues in building neural network models; network architecture and size of training samples.Experiments were designed and carried out on two-group classification problems to find answers to these model building questions. The first experiment deals with selection of architecture and sample size for different classification problems. Results show that choice of architecture and choice of sample size depend on the objective: to maximize the classification rate of training samples, or to maximize the generalizability of neural networks. The second experiment compares neural network models with classical models such as linear discriminant analysis and quadratic discriminant analysis, and nonparametric methods such as k-nearest-neighbor and linear programming. Results show that neural networks are comparable to, if not better than, these other methods in terms of classification rates in the training samples but not in the test samples.

Artificial neural networks in bankruptcy prediction: General framework and cross-validation analysis

“…Recently, Hung and Denton [27] and Subramanian and Hung [59] have proposed to use a general-purpose nonlinear optimizer, GRG2, in training neural networks. The bene®ts of GRG2 have been reported in the literature for many classi®cation problems [35,42,59]. This study uses a GRG2 based system to train neural networks.…”

“…If the objective is to classify a given set of observations in the training sample as well as possible, a larger network may be desirable. On the other hand, if the network is used to predict classi®cation of unseen objects in the test sample, then a larger network is not necessarily appropriate [42]. To see the eect of hidden nodes on the performance of neural network classi®ers, we use 15 dierent levels of hidden nodes ranging from 1 to 15 in this study.…”

“…Network architecture refers to the number of layers, nodes in each layer and the number of arcs. Based on the results from [14,23,37,42], networks with one hidden layer is generally sucient for most problems including classi®cation. All networks used in this study will have one hidden layer.…”

Prediction of surface location error in milling considering the effects of uncertain factors

“…Alternatively, using too many neurons would increase training time and/or result in ANN losing its generalization attribute. In this study, number of nodes in the hidden layer is chosen as n 2 = 2 × n 1 + 1 (Patuwo et al, 1993), where n 1 is number of input variables. Then, the output of the ANN can be written as:…”

Performance Analysis of Neural Networks Training using Real Coded Genetic Algorithm

“…Multilayer perceptron have been applied successfully to model classifier using backpropagation algorithm for approximating all linearly and non-linearly separable complex functions. Multilayer perceptron has popularity in solving several business and scientific problems that involve prediction, and has also a wide range of applications in the classification problems (Denton, Hung, & Osyk [7], Holmstrom, Koistinen, Laaksonen, & Oja [8]; Mangiameli West, [9]; Patwo, Hu, & Hung [10]; Pendharkar [11]). But MLP has major problem of getting stuck at the local optimal solutions during training and the weights of the network layers are updated through error backpropagation is time consuming.…”