• The Korean Journal of Applied Statistics
• /
• v.21 no.1
• /
• pp.109-123
• /
• 2008

Many small area estimation methods have been suggested. Also for the comparison of the estimation methods, model diagnostic checking techniques have been studied. Almost all of the small area estimators were developed by minimizing MSE(Mean square error) and so the MSE is the well-known comparison criterion for superiority. In this paper we suggested a new small area estimator based on minimizing MSPE(Mean square percentage error) which is recently re-highlighted. Also we compared the new suggested estimator with the estimators explained in Shin et al. (2007) using MSE, MSPE and other diagnostic checking criteria.

• The Korean Journal of Applied Statistics
• /
• v.26 no.6
• /
• pp.943-952
• /
• 2013

Wald, Agresti-Coull, Jeffreys, and Bayes-Laplace methods are commonly used for confidence interval of binomial proportion are applied for prediction intervals. We used coverage probability, mean coverage probability, root mean squared error, and mean expected width for numerical comparisons. From the comparisons, we found that Wald is not proper as for confidence interval and Agresti-Coull is too conservative to differ from confidence interval. However, Jeffrey and Bayes-Laplace are good for prediction interval and Jeffrey is especially desirable as for confidence interval.

• The Korean Journal of Applied Statistics
• /
• v.34 no.4
• /
• pp.579-588
• /
• 2021

Wald, Agresti-Coull, Jeffreys, and Bayes-Laplace methods are commonly used for confidence interval of binomial proportion are applied for prediction intervals. We used coverage probability, mean coverage probability, root mean squared error, and mean expected width for numerical comparisons. From the comparisons, we found that Wald is not proper as for confidence interval and Agresti-Coull is too conservative to differ from confidence interval. However, Jeffrey and Bayes-Laplace are good for prediction interval and Jeffrey is especially desirable as for confidence interval.

• Journal of Intelligence and Information Systems
• /
• v.5 no.1
• /
• pp.95-101
• /
• 1999

In this paper, a comparison between different methods to combine predictions from neural networks will be given. These methods are bagging, bumping, and balancing. Those are based on the analysis of the ensemble generalization error into an ambiguity term and a term incorporating generalization performances of individual networks. Neural Networks and AI machine learning models are prone to overfitting. A strategy to prevent a neural network from overfitting, is to stop training in early stage of the learning process. The complete data set is spilt up into a training set and a validation set. Training is stopped when the error on the validation set starts increasing. The stability of the networks is highly dependent on the division in training and validation set, and also on the random initial weights and the chosen minimization procedure. This causes early stopped networks to be rather unstable: a small change in the data or different initial conditions can produce large changes in the prediction. Therefore, it is advisable to apply the same procedure several times starting from different initial weights. This technique is often referred to as training ensembles of neural networks. In this paper, we presented a comparison of three statistical methods to prevent overfitting of neural network.

• Journal of the Optical Society of Korea
• /
• v.17 no.6
• /
• pp.525-532
• /
• 2013

Contrary to the academic interests of other existing studies elsewhere, this study deals with how the alignment algorithms such as sensitivity or Differential Wavefront Sampling (DWS) can be better used under effects from field, compensator positioning and environmental errors unavoidable from the shop-floor alignment work. First, the influences of aforementioned errors to the alignment state estimation was investigated with the algorithms. The environmental error was then found to be the dominant factor influencing the alignment state prediction accuracy. Having understood such relationship between the distorted system wavefront caused by the error sources and the alignment state prediction, we used it for simulated and experimental alignment runs for Infrared Optical System (IROS). The difference between trial alignment runs and experiment was quite close, independent of alignment methods; 6 nm rms for sensitivity method and 13 nm rms for DWS. This demonstrates the practical usefulness and importance of the prior error analysis using the alignment algorithms before the actual alignment runs begin. The error analysis methodology, its application to the actual alignment of IROS and their results are described together with their implications.

• Journal of the Korean Society for Precision Engineering
• /
• v.28 no.2
• /
• pp.168-175
• /
• 2011

Contouring accuracy of CNC machine tools is very important for high-speed and high-precision machining. In particular, large contour error may occur during corner tracking. In order to reduce the corner contouring error, acceleration and deceleration control or tool-path planning methods have been suggested. However, they do not directly control the corner contouring error. In the meantime, network servo systems are widely used because of their easiness of building and cost effectiveness. Communication latency between the master controller and servo drives, however, may deteriorate contouring accuracy especially during corner tracking. This paper proposes a control strategy that can accurately calculate and directly control the corner contouring error. A prediction control is combined with the above control to cope with communication latency. The proposed control method is evaluated through computer simulation and experiments. The results show its validity and usefulness.

• Communications for Statistical Applications and Methods
• /
• v.31 no.4
• /
• pp.441-457
• /
• 2024

Semicontinuous data are characterized by a mixture of a point probability mass at zero and a continuous distribution of positive values. This type of data is often modeled using a two-part model where the first part models the probability of dichotomous outcomes -zero or positive- and the second part models the distribution of positive values. Despite the two-part model's popularity, variable selection in this model has not been fully addressed, especially, in high dimensional data. The objective of this study is to investigate variable selection and prediction performance of penalized regression methods in two-part models. The performance of the selected techniques in the two-part model is evaluated via simulation studies. Our findings show that LASSO and ENET tend to select more predictors in the model than SCAD and MCP. Consequently, MCP and SCAD outperform LASSO and ENET for β-specificity, and LASSO and ENET perform better than MCP and SCAD with respect to the mean squared error. We find similar results when applying the penalized regression methods to the prediction of crime incidents using community-based data.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.95-101
• /
• 2008

Modeling of stock prices forecast has been considered as one of the most difficult problem to develop accurately since stock prices are highly correlated with various environmental conditions including economics and political situation. In this paper, we propose a agent system approach to predict Korea Composite Stock Price Index (KOSPI) using neural network and statistical methods. To minimize mean of prediction error and variation of prediction error, agent system includes sub-agent modules for feature extraction, variables selection, forecast engine selection, and forecasting results analysis. As a first step to develop agent system for KOSPI forecasting, twelve economic indices are selected from twenty two basic standard economic indices using principal component analysis. From selected twelve economic indices, prediction model input variables are chosen again using best-subsets regression method. Two different types data are tested for KOSPI forecasting and the Prediction results showed 11.92 points of root mean squared error for consecutive thirty days of prediction. Also, it is shown that proposed agent system approach for KOSPI forecast is effective since required types and numbers of prediction variables are time-varying, so adaptable selection of modeling inputs and prediction engine are essential for reliable and accurate forecast model.

• ETRI Journal
• /
• v.31 no.2
• /
• pp.162-172
• /
• 2009

A linear-prediction-based blind equalization algorithm for single-input single-output (SISO) finite impulse response/infinite impulse response (FIR/IIR) channels is proposed. The new algorithm is based on second-order statistics, and it does not require channel order estimation. By oversampling the channel output, the SISO channel model is converted to a special single-input multiple-output (SIMO) model. Two forward linear predictors with consecutive prediction delays are applied to the subchannel outputs of the SIMO model. It is demonstrated that the partial parameters of the SIMO model can be estimated from the difference between the prediction errors when the length of the predictors is sufficiently large. The sufficient filter length for achieving the optimal prediction is also derived. Based on the estimated parameters, both batch and adaptive minimum-mean-square-error equalizers are developed. The performance of the proposed equalizers is evaluated by computer simulations and compared with existing algorithms.

• Advances in nano research
• /
• v.14 no.3
• /
• pp.225-234
• /
• 2023

In this study, we aim to use big data resources and statistical analysis to obtain a reliable instruction to reach high-quality and high yield agricultural yields. In this regard, soil type data, raining and temperature data as well as wheat production in each year are collected for a specific region. Using statistical methodology, the acquired data was cleaned to remove incomplete and defective data. Afterwards, using several classification methods in machine learning we tried to distinguish between different factors and their influence on the final crop yields. Comparing the proposed models' prediction using statistical quantities correlation factor and mean squared error between predicted values of the crop yield and actual values the efficacy of machine learning methods is discussed. The results of the analysis show high accuracy of machine learning methods in the prediction of the crop yields. Moreover, it is indicated that the random forest (RF) classification approach provides best results among other classification methods utilized in this study.