• Journal of Digital Convergence
• /
• v.11 no.11
• /
• pp.409-414
• /
• 2013

Naive Bayesian classifiers are a powerful and well-known type of classifiers that can be easily induced from a dataset of sample cases. However, the strong conditional independence assumptions can sometimes lead to weak classification performance. Normally, naive Bayesian classifiers use Gaussian distributions to handle continuous attributes and to represent the likelihood of the features conditioned on the classes. The probability density of attributes, however, is not always well fitted by a Gaussian distribution. Another eminent type of classifier is the neuro-fuzzy classifier, which can learn fuzzy rules and fuzzy sets using supervised learning. Since there are specific structural similarities between a neuro-fuzzy classifier and a naive Bayesian classifier, the purpose of this study is to apply learning distribution graphs constructed by a neuro-fuzzy network to naive Bayesian classifiers. We compare the Gaussian distribution graphs with the fuzzy distribution graphs for the naive Bayesian classifier. We applied these two types of distribution graphs to classify leukemia and colon DNA microarray data sets. The results demonstrate that a naive Bayesian classifier with fuzzy distribution graphs is more reliable than that with Gaussian distribution graphs.

• Communications for Statistical Applications and Methods
• /
• v.27 no.1
• /
• pp.97-108
• /
• 2020

This paper compares several methods for estimating parameters of normal inverse Gaussian distribution. Ordinary maximum likelihood estimation and the method of moment estimation often do not work properly due to restrictions on parameters. We examine the performance of adjusted estimation methods along with the ordinary maximum likelihood estimation and the method of moment estimation by simulation and real data application. We also see the effect of the initial value in estimation methods. The simulation results show that the ordinary maximum likelihood estimator is significantly affected by the initial value; in addition, the adjusted estimators have smaller root mean square error than ordinary estimators as well as less impact on the initial value. With real datasets, we obtain similar results to what we see in simulation studies. Based on the results of simulation and real data application, we suggest using adjusted maximum likelihood estimates with adjusted method of moment estimates as initial values to estimate the parameters of normal inverse Gaussian distribution.

• Communications for Statistical Applications and Methods
• /
• v.2 no.2
• /
• pp.166-175
• /
• 1995

When a neutral particle beam(NPB) aimed at the object and receive a small number of neutron signals at the detector without any errors, it obeys Poisson law. Under the two assumptions that neutral particle scattering distribution and aiming errors have a circular Gaussian distributions that neutral particle scattering distribution and aiming errors have a circular Gaussian distribution respectively, an exact probability distribution of neutral particles vecomes a Poisson-power function distribution. We study and prove some properties, such as limiting distribution, unimodality, stochastical ordering, computational recursion fornula, of this distribution. We also prove monotone likelihood ratio(MLR) property of this distribution. Its MLR property can be used to find a criteria for the hypothesis testing problem.

• The Korean Journal of Applied Statistics
• /
• v.26 no.1
• /
• pp.37-47
• /
• 2013

A Q-Q plot is an effective and convenient graphical method to assess a distributional assumption of data. The primary step in the construction of a Q-Q plot is to obtain a closed-form expression to represent the relation between observed quantiles and theoretical quantiles to be plotted in order that the points fall near the line y = a + bx. In this paper, we introduce a Q-Q plot to assess goodness-of-fit for inverse Gaussian distribution. The procedure is based on the distributional result that a transformed random variable $Y={\mid}\sqrt{\lambda}(X-{\mu})/{\mu}\sqrt{X}{\mid}$ follows a half-normal distribution with mean 0 and variance 1 when a random variable X has an inverse Gaussian distribution with location parameter ${\mu}$ and scale parameter ${\lambda}$. Simulations are performed to provide a guideline to interpret the pattern of points on the proposed inverse Gaussian Q-Q plot. An illustrative example is provided to show the usefulness of the inverse Gaussian Q-Q plot.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.18 no.5
• /
• pp.1143-1148
• /
• 2014

This paper has analyzed the change of subthreshold swing for doping distribution function of asymmetric double gate(DG) MOSFET. The basic factors to determine the characteristics of DGMOSFET are dimensions of channel, i.e. channel length and channel thickness, and doping distribution function. The doping distributions are determined by ion implantation used for channel doping, and follow Gaussian distribution function. Gaussian function has been used as carrier distribution in solving the Poisson's equation. Since the Gaussian function is exactly not symmetric for top and bottome gates, the subthreshold swings are greatly changed for channel length and thickness, and the voltages of top and bottom gates for asymmetric double gate MOSFET. The deviation of subthreshold swings has been investigated for parameters of Gaussian distribution function such as projected range and standard projected deviation in this paper. As a result, we know the subthreshold swing is greatly changed for doping profiles and bias voltage.

• Wind and Structures
• /
• v.34 no.5
• /
• pp.421-435
• /
• 2022

To investigate the non-Gaussian properties of fluctuating wind pressures and the error margin of extreme wind loads on a long-span curved roof with matching and mismatching ratios of turbulence integral scales to depth (L_u^x/D), a series of synchronized pressure tests on the rigid model of the complex curved roof were conducted. The regions of Gaussian distribution and non-Gaussian distribution were identified by two criteria, which were based on the cumulative probabilities of higher-order statistical moments (skewness and kurtosis coefficients, S_k and K_u) and spatial correlation of fluctuating wind pressures, respectively. Then the characteristics of fluctuating wind-loads in the non-Gaussian region were analyzed in detail in order to understand the effects of turbulence integral-scale. Results showed that the fluctuating pressures with obvious negative-skewness appear in the area near the leading edge, which is categorized as the non-Gaussian region by both two identification criteria. Comparing with those in the wind field with matching L_u^x/D, the range of non-Gaussian region almost unchanged with a smaller L_u^x/D, while the non-Gaussian features become more evident, leading to higher values of S_k, Ku and peak factor. On contrary, the values of fluctuating pressures become lower in the wind field with a smaller L_u^x/D, eventually resulting in underestimation of extreme wind loads. Hence, the matching relationship of turbulence integral scale to depth should be carefully considered as estimating the extreme wind loads of long-span roof by wind tunnel tests.

• ETRI Journal
• /
• v.32 no.1
• /
• pp.145-147
• /
• 2010

In this letter, we present a new approximation for the twodimensional (2-D) Gaussian Q-function. The result is represented by only the one-dimensional (1-D) Gaussian Q-function. Unlike the previous 1-D Gaussian-type approximation, the presented approximation can be applied to compute the 2-D Gaussian Q-function with large correlations.

• Journal of the Korean Geophysical Society
• /
• v.5 no.2
• /
• pp.131-142
• /
• 2002

Geomagnetic transfer function is generally estimated by choosing transfer to minimize the square sum of differences between observed values. If the error structure sccords to the Gaussian distribution, standard least square(LS) can be the estimation. However, for non-Gaussian error distribution, the LS estimation can be severely biased and distorted. In this paper, the Gaussian error assumption was tested by Q-Q(Quantile-Quantile) plot which provided information of real error structure. Therefore, robust estimation such as regression M-estimate that does not allow a few bad points to dominate the estimate was applied for error structure with non-Gaussian distribution. The results indicate that the performance of robust estimation is similar to the one of LS estimation for Gaussian error distribution, whereas the robust estimation yields more reliable and smooth transfer function estimates than standard LS for non-Gaussian error distribution.

• The Korean Journal of Applied Statistics
• /
• v.29 no.4
• /
• pp.741-752
• /
• 2016

Numerous studies have shown that normal inverse Gaussian (NIG) distribution adequately fits the empirical return distribution of financial securities. The estimation of parameters can also be done relatively easily, which makes the NIG distribution more useful in financial markets. The maximum likelihood estimation and the method of moments estimation are easy to implement; however, we may encounter a problem in practice when a relationship among the moments is violated. In this paper, we investigate this problem in the parameter estimation and try to find a simple solution through simulations. We examine the effect of our adjusted estimation method with real data: daily log returns of KOSPI, S&P500, FTSE and HANG SENG. We also checked the performance of our method by computing the value at risk of daily log return data. The results show that our method improves the stability of parameter estimation, while it retains a comparable performance in goodness-of-fit.

• Ocean Systems Engineering
• /
• v.4 no.4
• /
• pp.293-308
• /
• 2014

Unlike the traditional displacement type vessels, the high speed planing crafts are supported by the lift forces which are highly non-linear. This non-linear phenomenon causes their motions in an irregular seaway to be non-Gaussian. In general, it may not be possible to express the probability distribution of such processes by an analytical formula. Also the process might not be stationary or ergodic in which case the statistical behavior of the motion to be constantly changing with time. Therefore the extreme values of such a process can no longer be calculated using the analytical formulae applicable to Gaussian processes. Since closed form analytical solutions do not exist, recourse is taken to fitting a distribution to the data and estimating the statistical properties of the process from this fitted probability distribution. The peaks over threshold analysis and fitting of the Generalized Pareto Distribution are explored in this paper as an alternative to Weibull, Generalized Gamma and Rayleigh distributions in predicting the short term extreme value of a random process.