• Journal of the Korean Statistical Society
• /
• v.27 no.4
• /
• pp.471-483
• /
• 1998

The influence of observations on the misclassification probability in multiple discriminant analysis under the equal covariance assumption is investigated by the local influence method. Under an appropriate perturbation we can get information about influential observations and outliers by studying the curvatures and the associated direction vectors of the perturbation-formed surface of the misclassification probability. We show that the influence function method gives essentially the same information as the direction vector of the maximum slope. An illustrative example is given for the effectiveness of the local influence method.

• Journal of the Korean Statistical Society
• /
• v.25 no.1
• /
• pp.145-151
• /
• 1996

The local behaviour of the surface formed by the perturbed maximum likelihood estimator of the squared Mahalanobis distance is investigated. The study of the local behaviour allows a simultaneous perturbation on the samples of interest and it is effective in identifying influential observations.

• ETRI Journal
• /
• v.16 no.1
• /
• pp.35-43
• /
• 1994

In this paper, the robustness of the artificial neural networks to noise is demonstrated with a multilayer perceptron, and the reason of robustness is due to the statistical orthogonality among hidden nodes and its hierarchical information extraction capability. Also, the misclassification probability of a well-trained multilayer perceptron is derived without any linear approximations when the inputs are contaminated with random noises. The misclassification probability for a noisy pattern is shown to be a function of the input pattern, noise variances, the weight matrices, and the nonlinear transformations. The result is verified with a handwritten digit recognition problem, which shows better result than that using linear approximations.

• Journal of the Korea Safety Management & Science
• /
• v.11 no.3
• /
• pp.189-198
• /
• 2009

This thesis studies two imputation methods, the MCMC method and the EM algorithm, that take care of the problem. The performance of the two methods for the linear (or quadratic) discriminant analysis are evaluated under various types of incomplete observations. Based on simulated experiments, the effect of the imputation using the EM algorithm and the MCMC method are evaluated and compared in terms of the probability of misclassification and the RMSE. This is done for the various cases of incomplete observations. The cases are differentiated by missing rates, sample sizes, and distances between two classification groups. The studies show that the probability of misclassification and the RMSE of the EM algorithm method is lower than the MCMC method. Therefore the imputation using the EM algorithm is more efficient than the MCMC method. And the probability of misclassification of the method that all vectors of observations with missing values are omitted from analysis is lower than the EM algorithm and the MCMC method when the samples size is small and the rate of missing values is extremely big.

• Journal of the Korean Data and Information Science Society
• /
• v.22 no.2
• /
• pp.149-157
• /
• 2011

Undecided inference could be regarded as a missing data problem such as MARand MNAR. Under the assumption of MAR, undecided inference make use of logistic regression model. The probability of default for the undecided group is obtained with regression coefficient vectors for the decided group and compare with the probability of default for the decided group. And under the assumption of MNAR, undecide dinference make use of logistic regression model with additional feature random vector. Simulation results based on two kinds of real data are obtained and compared. It is found that the misclassification rates are not much different from the rate of rawdata under the assumption of MAR. However the misclassification rates under the assumption of MNAR are less than those under the assumption of MAR, and as the ratio of the undecided group is increasing, the misclassification rates is decreasing.

• The Korean Journal of Applied Statistics
• /
• v.24 no.5
• /
• pp.941-949
• /
• 2011

The influence function is used to develop criteria to detect outliers in discriminant analysis. We derive the influence function of observations that estimate the the misclassification probability in discriminant analysis for three groups. The proposed measures are applied to the facial image data to define outliers and redo the discriminant analysis excluding the outliers. The study proves that the derived influence function is more efficient than using the discriminant probability approach.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.2
• /
• pp.71-79
• /
• 1997

In this paper, we propose a new nearest-neighbor classifier with reduced computational complexity in search process. In the proposed classifier, the classes are divided into two sets: reference and non-reference sets. It reduces computational requriement by approximating the distance between the input and a class iwth the information of distances among the calsses. It calculates only the distance between the input and the reference classes. We convert a given classifier into RCC (reduced computational complexity but smal lincrease in misclassification probability of its corresponding RCC classifier. We designed RCC classifiers for the recognition of digits from the NIST database. We obtained an RCC classifier with 60% reduction in the computational complexity with the cost of 0.5% increase in misclassification probability.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.13 no.1
• /
• pp.55-72
• /
• 2009

A misclassified size-biased modified power series distribution (MSBMPSD) where some of the observations corresponding to x = c + 1 are misclassified as x = c with probability $\alpha$, is defined. We obtain its recurrence relations among the raw moments, the central moments and the factorial moments. Discussion of the effect of the misclassification on the variance is considered. To illustrate the situation under consideration some of its particular cases like the size-biased generalized negative binomial (SBGNB), the size-biased generalized Poisson (SBGP) and sizebiased Borel distributions are included. Finally, an example is presented for the size-biased generalized Poisson distribution to illustrate the results.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.6
• /
• pp.1004-1015
• /
• 1994

Modulation type classifier based on statistical moments has been successfully employed to classify PSK signals. Previously, the classifier developed utilizes the statistical moment of samples of the received signal phase, which may be difficult to extract from received signal. In this paper we propose a new moments-based classifier to classify PSK signals by using the moments of the demodulated signal for PSK. THe demodulated signal can be easily extracted from the conventional demodulation of PSK. The evaluation of the performance of the proposed classifier for PSK signals has been investigated in additive white Gaussian noise environment using the exact distribution of the demodulated signal. The performances of classifier in terms of probability of misclassification were evaluated. We found that the coherent system classifier gave 4dB improvement for BPSK and 3dB for QPSK over noncoherent system classifier, when the probability of misclassification is 10 and m equals to 4.

• Journal of the Korean Statistical Society
• /
• v.24 no.2
• /
• pp.453-470
• /
• 1995

We describe a constrained optimal classification rule for the case when the prior probability of an observation belonging to one of the two populations is unknown. This is done by suggesting a balanced design for the classification experiment and constructing the optimal rule under the balanced design condition. The rule si characterized by a constrained minimization of total risk of misclassification; the constraint of the rule is constructed by the process of equation between Kullback-Leibler's directed divergence measures obtained from the two population conditional densities. The efficacy of the suggested rule is examined through two-group normal classification. This indicates that, in case little is known about the relative population sizes, dramatic gains in accuracy of classification result can be achieved.