• 한국ITS학회:학술대회논문집
• /
• 2003.11a
• /
• pp.297-302
• /
• 2003

• Proceedings of the Korean Society of Soil and Groundwater Environment Conference
• /
• 2000.05a
• /
• pp.65-65
• /
• 2000

• The Transactions of the Korea Information Processing Society
• /
• v.5 no.10
• /
• pp.2609-2614
• /
• 1998

In this work, we consider the forecasting performance between nearal network and Box-Jenkins method for time series data. A modified learning process is developed for neural network approach at time eries data, ie, properly adaptive learning rates selecting by orthogonal arrays and dynamic selecting of initial values using Easton's cotroller box. We can obtain good starting points with dynamic graphics approach. We use real data sets for this study : the Wolf yearly sunspot numbers between 1700 and 1988.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 1999.03a
• /
• pp.271-280
• /
• 1999

인공신경망을 시계열예측에 적용하는 경우에 고려되어야 할 문제중, 특히 모형에 적합한 입력변수의 생성이 중요시되고 있는데, 이러한 분야는 인공신경망의 모형생성과정에서 입력변수에 대한 전처리기법으로써 다양하게 제시되어 왔다. 가장 최근의 입력변수 전처리기법으로써 제시되고 있는 신호처리기법은 전통적 주기분할처리방법인 푸리에변환기법(Fourier transforms)을 비롯하여 이를 확장시킨 개념인 웨이블릿변환기법(wavelet transforms) 등으로 대별될 수 있다. 이는 기본적으로 시계열이 다수의 주기(cycle)들로 구성된 상이한 시계열들의 집합이라는 가정에서 출발하고 있다. 전통적으로 이러한 시계열은 전기 또는 전자공학에서 주파수영역분할, 즉 고주파 및 저주파수를 분할하기 위한 기법에 적용되어 왔다. 그러나, 최근에는 이러한 연구가 다양한 분야에 활발하게 응용되기 시작하였으며, 그 중의 대표적인 예가 바로 경영분야의 재무시계열에 대한 분석이다. 전통적으로 재무시계열은 장, 단기의사결정을 가진 시장참여자들간의 거래특성이 시계열에 각기 달리 가격으로 반영되기 때문에 이러한 상이한 집단들의 고요한 거래움직임으로 말미암아 예를 들어, 주식시장이 프랙탈구조를 가지고 있다고 보기도 한다. 이처럼 재무시계열은 다양한 사회현상의 집합체라고 볼 수 있으며, 그만큼 예측모형을 구축하는데 어려움이 따른다. 본 연구는 이러한 시계열의 주기적 특성에 기반을 둔 신호처리분석으로서 기존의 시계열로부터 노이즈를 줄여 주면서 보다 의미있는 정보로 변환시켜줄 수 있는 웨이블릿분석 방법론을 새로운 필터링기법으로 사용하여 현재 많은 연구가 진행되고 있는 인공신경망의 모형결합을 통해 기존연구과는 다른 새로운 통합예측방법론을 제시하고자 한다. 본 연구에서는 제시하는 통합방법론은 크게 2단계 과정을 거쳐 예측모형으로 완성이 된다. 즉, 1차 모형단계에서 원시 재무시계열은 먼저 웨이브릿분석을 통해서 노이즈가 필터링 되는 동시에, 과거 재무시계열의 프랙탈 구조, 즉 비선형적인 움직임을 보다 잘 반영시켜 주는 다차원 주기요소를 가지는 시계열로 분해, 생성되며, 이렇게 주기에 따라 장단기로 분할된 시계열들은 2차 모형단계에서 신경망의 새로운 입력변수로서 사용되어 최종적인 인공 신경망모델을 구축하는 데 반영된다. 기존의 주기분할방법론은 모형개발자입장에서 여러 가지 통계기준치중에서 최적의 기준치를 합리적으로 선택해야 하는 문제가 추가적으로 발생하며, 본 연구에서는 이상의 제반 문제들을 개선시키기 위해 통합방법론으로서 기존의 인공신경망모형을 구조적으로 확장시켰다. 이 모형에서 기존의 입력층 이전단계에 새로운 층이 정의된다. 이렇게 해서 생성된 새로운 통합모형은 기존모형에서 생성되는 기본적인 학습파라미터와 더불어, 본 연구에서 새롭게 제시된 주기분할층의 파라미터들이 모형의 학습성과를 높이기 위해 함께 고려된다. 한편, 이러한 학습과정에서 추가적으로 고려해야 할 파라미터 갯수가 증가함에 따라서, 본 모델의 학습성과가 local minimum에 빠지는 문제점이 발생될 수 있다. 즉, 웨이블릿분석과 인공신경망모형을 모두 전역적으로 최적화시켜야 하는 문제가 발생한다. 본 연구에서는 이 문제를 해결하기 위해서, 최근 local minimum의 가능성을 최소화하여 전역적인 학습성과를 높여 주는 인공지능기법으로서 유전자알고리즘기법을 본 연구이 통합모델에 반영하였다. 이에 대한 실증사례 분석결과는 일일 환율예측문제를 적용하였을 경우, 기존의 방법론보다 더 나운 예측성과를 타나내었다.

• Spatial Information Research
• /
• v.14 no.1 s.36
• /
• pp.163-176
• /
• 2006

The concept of 'spatial association' explains spatial distribution pattern of geographical phenomenon based on similarity with neighborhoods, as in the Tobler's Law of Geography: 'Everything is related to everything else, but near things are more related than distant things.' In this study, we develop a time-series exploratory analysis tool for discovering temporal patterns of spatial association by combining spatial statistics and geo-visualization, and thus present a possibility to support spatial decision-making process. As for the spatial proximity weight matrix indispensable to measuring global and local spatial association, we employ a variety of flexible weighting schemes using geometric characteristics of areal unit. In addition, we renovate the existing visualization methods for more effective understanding of the procedures and results of time-series analysis on spatial association: for instance, temporal parallel coordinate plot with box plot, animated map for spatial association, and 3D Moran scatterplot. The feasibility of our system is verified by time-series analysis experiments on the spatial association of land price fluctuation rate for all administrative units in Korea, $1995{\sim}2004$.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.1
• /
• pp.11-20
• /
• 2017

In this paper, we introduced linear spatial time series (space-time autoregressive and moving average model) and nonlinear spatial time series (space-time bilinear model). Also we estimated the parameters by Kalman Filter method and made comparative studies of power of forecast in the final model. We proposed several weight matrices such as equal proportion allocation, reciprocal proportion between distances, and proportion of population sizes. For applications, we collected Mumps data at Korea Center for Disease Control and Prevention from January 2001 until August 2008. We compared three approaches of weight matrices using the Mumps data. Finally, we also decided the most effective model based on sum of square forecast error.

• Asia-Pacific Journal of Business Venturing and Entrepreneurship
• /
• v.9 no.2
• /
• pp.15-21
• /
• 2014

The long memory properties of the hedge ratio for stock and futures have not been systematically investigated by the extant literature. To investigate hedge ratio' long memory, this paper employs a data set including KOSPI200 and S&P500. Coakley, Dollery, and Kellard(2008) employ a data set including a stock index and commodities foreign exchange, and suggested the S&P500 to be a fractionally integrated process. This paper firstly estimates hedge ratios with two dynamic models, BEKK(Bollerslev, Engle, Kroner, and Kraft) and diagonal-BEKK, and tests the long memory of hedge ratios with Geweke and Porter-Hudak(1983)(henceforth GPH) and Lo's modified rescaled adjusted range test by Lo(1991). In empirical results, two hedge ratios based on KOSPI200 and S&P500 show considerably significant long memory behaviours. Thus, such results show the hedge ratios to be stationary and strongly reject the random walk hypothesis on hedge ratios, which violates the efficient market hypothesis.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.12
• /
• pp.11-21
• /
• 2011

An embedded sensor is significantly influenced by its spatial environment, such as barriers or distance, through low power and signal strength. Due to these causes, noise data frequently occur in an embedded sensor. Because the information acquired from the embedded sensor exists in a time series, it is hard to detect an error which continuously takes place in the time series information on a realtime basis. In this paper, we proposes an error detection method based on time-series prediction that detects error signals of embedded sensors in real time in consideration of the physical characteristics of embedded devices. The error detection method based on time-series prediction proposed in this paper determines errors in generated embedded device signals using a stable distance function. When detecting errors by monitoring signals from an embedded device, the stable distance function can detect error signals effectively by applying error weight to the latest signals. When detecting errors by monitoring signals from an embedded device, the stable distance function can detect error signals effectively by applying error weight to the latest signals.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.28 no.3
• /
• pp.361-367
• /
• 2010

We produced continuous vertical time series of 14 permanent GPS stations operated by National Geographic Information Institute by processing about five years of data. Then we computed the height velocities by using a linear regression fitting of those time series, and did principal component analysis to understand the overall characteristics of the series. The prominent signal obtained as the first mode of PCA results showed an average of 4.2 mm/yr vertical velocity. The values of the first mode eigenvectors were consistent at all sites. Thus, we concluded that all the 14 stations are uplifting nearly at the same velocity for the test period. Then changes of precision before and after removing the first mode signal from the 14 height time series were analyzed. As a result, the precision improved 34.8% on average.

• Journal of Intelligence and Information Systems
• /
• v.19 no.1
• /
• pp.1-17
• /
• 2013

In this article, several types of hybrid forecasting models are suggested. In particular, hybrid models using the generalized additive model (GAM) are newly suggested as an alternative to those using neural networks (NN). The prediction performances of various hybrid and non-hybrid models are evaluated using simulated time series data. Five different types of seasonal time series data related to an additive or multiplicative trend are generated over different levels of noise, and applied to the forecasting evaluation. For the simulated data with only seasonality, the autoregressive (AR) model and the hybrid AR-AR model performed equivalently very well. On the other hand, if the time series data employed a trend, the SARIMA model and some hybrid SARIMA models equivalently outperformed the others. In the comparison of GAMs and NNs, regarding the seasonal additive trend data, the SARIMA-GAM evenly performed well across the full range of noise variation, whereas the SARIMA-NN showed good performance only when the noise level was trivial.