• 산업경영시스템학회지
• /
• 제12권19호
• /
• pp.31-37
• /
• 1989

본 논문은 비선형 회귀분석 모델(Nonlinear Regression Models)에서의 추산잔차(Recursive Residuals)를 정의하기 위한 것을 목적으로 한다. 선형 회귀분석 모델(Linear Regression Models)에서는 추산잔차가 우리가 측정할 수 없는 진짜 오차(True Error)와 같은 확률 분포를 갖는데 이의 평균은 0이고 분산은 ${sigma}^2$이다. 그러나 비선형 회귀분석에서는 이와 같은 정확한 분포를 알 수가 없기 때문에, 여러 종류의 잔차들을 연구 검토하고 나아가서 시뮬레이션(Simulation)을 통하여 분석.비교한 뒤 추산잔차를 정의하기로 한다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제13권3호
• /
• pp.186-199
• /
• 2013

We pose pattern classification as a density estimation problem where we consider mixtures of generative models under partially labeled data setups. Unlike traditional approaches that estimate density everywhere in data space, we focus on the density along the decision boundary that can yield more discriminative models with superior classification performance. We extend our earlier work on the recursive estimation method for discriminative mixture models to semi-supervised learning setups where some of the data points lack class labels. Our model exploits the mixture structure in the functional gradient framework: it searches for the base mixture component model in a greedy fashion, maximizing the conditional class likelihoods for the labeled data and at the same time minimizing the uncertainty of class label prediction for unlabeled data points. The objective can be effectively imposed as individual mixture component learning on weighted data, hence our mixture learning typically becomes highly efficient for popular base generative models like Gaussians or hidden Markov models. Moreover, apart from the expectation-maximization algorithm, the proposed recursive estimation has several advantages including the lack of need for a pre-determined mixture order and robustness to the choice of initial parameters. We demonstrate the benefits of the proposed approach on a comprehensive set of evaluations consisting of diverse time-series classification problems in semi-supervised scenarios.

• Journal of the Korean Statistical Society
• /
• 제30권2호
• /
• pp.281-290
• /
• 2001

Symmetric estimation and recursive mean adjustment are considered to construct tests for the doble unit root hypothesis for both parametric and semiparametric time series models. It is shown that simultaneous application of symmetric estimation and recursive mean adjustment yields the most powerful test. Moreover, size property of the semiparametric test based on the simultaneous application is bet among all semiparametric tests.

• STI Policy Review
• /
• 제2권2호
• /
• pp.45-54
• /
• 2011

This paper develops a recursive multinomial probit model and describes its estimation method. The recursive multinomial probit model is an extension of a recursive bivariate probit model. The main difference between the two models is that a single decision among two or more alternatives can be considered in each choice equation in the proposed model. The recursive multinomial probit model is developed based on a standard framework of the multinomial probit model and a Bayesian approach with a Gibbs sampling is adopted for the estimation. The simulation exercise with artificial data sets is showed that the model performed well. Since the recursive multinomial probit model can be applied to analyze the causal relationship between discrete dependent variables with more than two outcomes, the model can play an important role in extending the methodology of the causal relationship analysis in innovation research.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1986년도 하계학술대회논문집
• /
• pp.140-144
• /
• 1986

In this paper an FIR(finite impulse response) filter and smoother are introduced for discrete time-invariant state-space models with driving noises. The FIR structure not only quarantees the BIBO stability and the robustness to parameter changes but also improves the filter divergence problem. It is shown that the impulse responses of the FIR filter and the smoother are obtained by Riccati-type difference equations and that they are to be time-invariant and reduced to very simple forms. For implementational purpose, recursive forms of the FIR filler and smoother are derived with each other used as the adjoint variable.

• 무역학회지
• /
• 제46권6호
• /
• pp.191-209
• /
• 2021

In this paper, various time series analysis models and machine learning models are presented for long-term prediction of export growth rate, and the prediction performance is compared and reviewed by RMSE and MAE. Export growth rate is one of the major economic indicators to evaluate the economic status. And It is also used to predict economic forecast. The export growth rate may have a negative (-) value as well as a positive (+) value. Therefore, Instead of using the ReLU function, which is often used for time series prediction of deep learning models, the PReLU function, which can have a negative (-) value as an output value, was used as the activation function of deep learning models. The time series prediction performance of each model for three types of data was compared and reviewed. The forecast data of long-term prediction of export growth rate was deduced by three forecast methods such as a fixed forecast method, a recursive forecast method and a rolling forecast method. As a result of the forecast, the traditional time series analysis model, ARDL, showed excellent performance, but as the time period of learning data increases, the performance of machine learning models including LSTM was relatively improved.

• 한국신뢰성학회:학술대회논문집
• /
• 한국신뢰성학회 2002년도 정기학술대회
• /
• pp.263-272
• /
• 2002

Reliability of multi-phased mission system is represented where redundant components are repairable. Failures and repairs of components follow Markovian property Under some constraints, 4 models are available. Two models are represented here. The solutions are obtained as recursive equations using Markov model and eigenvalue system.

• Journal of the Korean Data and Information Science Society
• /
• 제8권2호
• /
• pp.271-276
• /
• 1997

A dynamic discount approach is proposed for the estimation of the Poisson parameter and the forecasting of the Poisson random variable, where the parameter of the Poisson distribution varies over time intervals. The recursive estimation procedure of the Poisson parameter is provided. Also the forecasted distribution of the Poisson random variable in the next time interval based on the information gathered until the current time interval is provided.

• Journal of the Korean Data and Information Science Society
• /
• 제7권2호
• /
• pp.227-234
• /
• 1996

The discount survival model is proposed for the application of the Cox model on the analysis of survival data with time-varying effects of covariates. Algorithms for the recursive estimation of the parameter vector and the retrospective estimation of the survival function are suggested. Also the algorithm of forecasting of the survival function of individuals of specific covariates in the next time interval based on the information gathered until the end of a certain time interval is suggested.

• Journal of the Korean Statistical Society
• /
• 제25권4호
• /
• pp.567-576
• /
• 1996

The multiprocess dynamic survival model is proposed for the application of the regression model on the analysis of survival data with time-varying effects of covariates : where the survival data consists of numbers of deaths at certain time-points. The algorithm for the recursive estimation of a time-varying parameter vector is suggested. Also the algorithm of forecasting of numbers of deaths of each group in the next time interval based on the information gathered until the end of current time interval is suggested.