• Journal of the Korean Data and Information Science Society
• /
• 제14권3호
• /
• pp.641-651
• /
• 2003

This study suggests the hybrid models for interest rate forecasting using structural changes (or change points). The basic concept of this proposed model is to obtain significant intervals caused by change points, to identify them as the change-point groups, and to reflect them in interest rate forecasting. The model is composed of three phases. The first phase is to detect successive structural changes in the U. S. Treasury bill rate dataset. The second phase is to forecast the change-point groups with data mining classifiers. The final phase is to forecast interest rates with backpropagation neural networks (BPN). Based on this structure, we propose three hybrid models in terms of data mining classifier: (1) multivariate discriminant analysis (MDA)-supported model, (2) case-based reasoning (CBR)-supported model, and (3) BPN-supported model. Subsequently, we compare these models with a neural network model alone and, in addition, determine which of three classifiers (MDA, CBR and BPN) can perform better. For interest rate forecasting, this study then examines the prediction ability of hybrid models to reflect the structural change.

• 한국농공학회지
• /
• 제45권2호
• /
• pp.45-57
• /
• 2003

An artificial neural network model was developed to analyze and forecast Short-term river runoff from the Naju watershed, in Korea. Error back propagation neural networks (EBPN) of hourly rainfall and runoff data were found to have a high performance In forecasting runoff. The number of hidden nodes were optimized using total error and Bayesian information criterion. Model forecasts are very accurate (i.e., relative error is less than 3% and $R^2$is greater than 0.99) for calibration and verification data sets. Increasing the time horizon for application data sets, thus mating the model suitable for flood forecasting. decreases the accuracy of the model. The resulting optimal EBPN models for forecasting hourly runoff consists of ten rainfall and four runoff data(ANN0410 model) and ten rainfall and ten runoff data(ANN1010 model). Performances of the ANN0410 and ANN1010 models remain satisfactory up to 6 hours (i.e., $R^2$is greater than 0.92).

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2008년도 학술발표회 논문집
• /
• pp.2193-2196
• /
• 2008

In this study, the new methodology such as support vector machines neural networks model (SVM-NNM) using the statistical learning theory is introduced to forecast flood stage in Nakdong river, Republic of Korea. The SVM-NNM in hydrologic time series forecasting is relatively new, and it is more problematic in comparison with classification. And, the multilayer perceptron neural networks model (MLP-NNM) is introduced as the reference neural networks model to compare the performance of SVM-NNM. And, for the performances of the neural networks models, they are composed of training, cross validation, and testing data, respectively. From this research, we evaluate the impact of the SVM-NNM and the MLP-NNM for the forecasting of the hydrologic time series in Nakdong river. Furthermore, we can suggest the new methodology to forecast the flood stage and construct the optimal forecasting system in Nakdong river, Republic of Korea.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2006년도 심포지엄 논문집 정보 및 제어부문
• /
• pp.270-272
• /
• 2006

In this paper, we discuss optimal design of Fuzzy Polynomial Neural Networks by means of Genetic Algorithms(GAs) using symbolic coding for non-linear data. One of the major subject of genetic algorithms is representation of chromosomes. The proposed model optimized by the means genetic algorithms which used symbolic code to represent chromosomes. The proposed gFPNN used a triangle and a Gaussian-like membership function in premise part of rules and design the consequent structure by constant and regression polynomial (linear, quadratic and modified quadratic) function between input and output variables. The performance of the proposed model is quantified through experimentation that exploits standard data already used in fuzzy modeling. These results reveal superiority of the proposed networks over the existing fuzzy and neural models.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2000년도 하계학술대회 논문집 D
• /
• pp.2946-2948
• /
• 2000

In this paper, we proposed the Fuzzy Polynomial Neural Networks(FPNN) model with fuzzy activation node. The proposed FPNN structure is generated from the mutual combination of PNN(Polynomial Neural Networks) structure and fuzzy inference system. The premise of fuzzy inference rules defines by triangular and gaussian type membership function. The fuzzy inference method uses simplified and regression polynomial inference method which is based on the consequence of fuzzy rule expressed with a polynomial such as linear, quadratic and modified quadratic equation are used. The structure of FPNN is not fixed like in conventional Neural Networks and can be generated. The design procedure to obtain an optimal model structure utilizing FPNN algorithm is shown in each stage. Gas furnace time series data used to evaluate the performance of our proposed model.

• International Journal of Quality Innovation
• /
• 제6권1호
• /
• pp.109-118
• /
• 2005

The parameter design is the most emphasized measure by researchers for a new products development. It is critical for makers to achieve simultaneously in both the time-to-market production and the quality enhancement. However, there are difficulties in practical application, such as (1) complexity and nonlinear relationships co-existed among the system's inputs, outputs and control parameters, (2) interactions occurred among parameters, (3) where the adjustment factors of Taguchi's two-phase optimization procedure cannot be sure to exist in practice, and (4) for some reasons, the data became lost or were never available. For these incomplete data, the Taguchi methods cannot treat them well. Neural networks have a learning capability of fault tolerance and model free characteristics. These characteristics support the neural networks as a competitive tool in processing multivariable input-output implementation. The successful fields include diagnostics, robotics, scheduling, decision-making, prediction, etc. This research is a case study of spherical annealing model. In the beginning, an original model is used to pre-fix a model of parameter design. Then neural networks are introduced to achieve another model. Study results showed both of them could perform the highest spherical level of quality.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2009년도 학술발표회 초록집
• /
• pp.112-115
• /
• 2009

The research of climate change impact in hydrometeorology often relies on climate change information. In this paper, neural networks models such as support vector machine neural networks model (SVM-NNM) and multilayer perceptron neural networks model (MLP-NNM) are proposed statistical downscaling of the monthly precipitation. The input nodes of neural networks models consist of the atmospheric meteorology and the atmospheric pressure data for 2 grid points including $127.5^{\circ}E/35^{\circ}N$ and $125^{\circ}E/35^{\circ}N$, which produced the best results from the previous study. The output node of neural networks models consist of the monthly precipitation data for Seoul station. For the performances of the neural networks models, they are composed of training and test performances, respectively. From this research, we evaluate the impact of SVM-NNM and MLP-NNM performances for the downscaling of the monthly precipitation data. We should, therefore, construct the credible monthly precipitation data for Seoul station using statistical downscaling method. The proposed methods can be applied to future climate prediction/projection using the various climate change scenarios such as GCMs and RCMs.

• 한국전산구조공학회:학술대회논문집
• /
• 한국전산구조공학회 1998년도 가을 학술발표회 논문집
• /
• pp.447-454
• /
• 1998

The paper describes the second half of the research for the development of Neural-Networks-based model for the generation of an Artificial earthquake and a Response Spectrum(NNARS). Based on the redefined traditional processes related to the generation of an earthquake acceleration response spectrum and design spectrum, four neural-networks-based models are proposed to substitute the traditional processes. RS_NN tries to directly generate acceleration response spectrum with basic data that are magnitude, epicentral distance, site conditions and focal depth. The test results of RS_NN are not good because of the characteristics of white noise, which is randomly generated. ARS_NN solve this problem by the introduction of the average concept. IARS_NN has a role to inverse the ARS_NN, so that is applied to generate a ground motion accelerogram compatible with the shape of a response spectrum. Additionally, DS_NN directly produces design spectrum with basic data. As these four neural networks are simulated as a step by step, the paper describes the methods to generate a response spectrum and a design spectrum using the neural networks.

• 스마트미디어저널
• /
• 제11권3호
• /
• pp.26-30
• /
• 2022

스파이킹 신경망은 실제 두뇌 뉴런의 작동원리를 적용한 신경망으로, 뉴런의 생물학적 메커니즘으로 인해 기존 신경망보다 학습과 추론에 소모되는 전력이 적다. 최근 딥러닝 모델이 거대해지며 운용에 소모되는 비용 또한 기하급수적으로 증가함에 따라 스파이킹 신경망은 합성곱, 순환 신경망을 잇는 3세대 신경망으로 주목받으며 관련 연구가 활발히 진행되고 있다. 그러나 스파이킹 신경망 모델을 산업에 적용하기 위해서는 아직 선행되어야 할 연구가 많이 남아있고, 새로운 모델을 적용하기 위한 모델 재학습 문제 역시 해결해야 한다. 본 논문에서는 기존의 학습된 딥러닝 모델의 가중치를 추출하여 스파이킹 신경망 모델의 가중치로 변환하는 것으로 모델 재학습 비용을 최소화하는 방법을 제안한다. 또한, 변환된 가중치를 사용한 추론 결과와 기존 모델의 결과를 비교해 가중치 변환이 올바르게 작동함을 보인다.

• 한국경영과학회:학술대회논문집
• /
• 한국경영과학회 1996년도 추계학술대회발표논문집; 고려대학교, 서울; 26 Oct. 1996
• /
• pp.31-34
• /
• 1996

The extensive researches have compared the performance of neural networks(NN) with those of various statistical techniques for the classification problem. The empirical results of these comparative studies have indicated that the neural networks often outperform the traditional statistical techniques. Moreover, there are some efforts that try to combine various classification methods, especially multivariate discriminant analysis with neural networks. While these efforts improve the performance, there exists a problem violating robust assumptions of multivariate discriminant analysis that are multivariate normality of the independent variables and equality of variance-covariance matrices in each of the groups. On the contrary, cluster analysis alleviates this assumption like neural networks. We propose a new approach to classification problems by combining the cluster analysis with neural networks. The resulting predictions of the composite model are more accurate than each individual technique.