• The Korean Journal of Applied Statistics
• /
• v.13 no.2
• /
• pp.525-540
• /
• 2000

Characteristics of precipitation in Seoul have been examined by using long-term observational data. Precipitation records from modern rain gauges were used for 1908-1996, together with the traditional Korean rain gauge (called Chukwookee) observations for 1777-1907. A linear trend analysis of seasonal total rainfall shows no significant trends over the last 200 years A wavelet transform analysis was performed to figure out the transient variations of precipitation.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_1
• /
• pp.587-596
• /
• 2022

Sea levels are rising as a result of climate change, and low-lying areas along the coast are at risk of flooding. Therefore, we tried to investigate the relationship between sea level change and climate indices using satellite altimeter data (Topex/Poseidon, Jason-1/2/3) and southern oscillation index (SOI) and the Pacific decadal oscillation (PDO) data. If time domain data were converted to frequency domain, the original data can be analyzed in terms of the periodic components. Fourier transform and Wavelet transform are representative periodic analysis methods. Fourier transform can provide only the periodic signals, whereas wavelet transform can obtain both the periodic signals and their corresponding time location. The cross-wavelet transformation and the wavelet coherence are ideal for analyzing the common periods, correlation and phase difference for two time domain datasets. Our cross-wavelet transform analysis shows that two climate indices (SOI, PDO) and sea level height was a significant in 1-year period. PDO and sea level height were anti-phase. Also, our wavelet coherence analysis reveals when sea level height and climate indices were correlated in short (less than one year) and long periods, which did not appear in the cross wavelet transform. The two wavelet analyses provide the frequency domains of two different time domain datasets but also characterize the periodic components and relative phase difference. Therefore, our research results demonstrates that the wavelet analyses are useful to analyze the periodic component of climatic data and monitor the various oceanic phenomena that are difficult to find in time series analysis.

• The Korean Journal of Applied Statistics
• /
• v.22 no.6
• /
• pp.1203-1214
• /
• 2009

In this paper we introduce the time series clustering methods in the time and frequency domains and discuss the merits or demerits of each method. We analyze 15 daily stock prices of KOSPI 200, and the nonparametric method using the wavelet shows the best clustering results. For the clustering of nonstationary time series using the spectral density, the EMD method remove the trend more effectively than the differencing.

• Smart Structures and Systems
• /
• v.22 no.2
• /
• pp.175-183
• /
• 2018

Time-series data often contain one of the most valuable pieces of information in many fields including manufacturing. Because time-series data are relatively cheap to acquire, they (e.g., vibration signals) have become a crucial part of big data even in manufacturing shop floors. Recently, deep-learning models have shown state-of-art performance for analyzing big data because of their sophisticated structures and considerable computational power. Traditional models for a machinery-monitoring system have highly relied on features selected by human experts. In addition, the representational power of such models fails as the data distribution becomes complicated. On the other hand, deep-learning models automatically select highly abstracted features during the optimization process, and their representational power is better than that of traditional neural network models. However, the applicability of deep-learning models to the field of prognostics and health management (PHM) has not been well investigated yet. This study integrates the "residual fitting" mechanism inherently embedded in the wavelet transform into the convolutional neural network deep-learning structure. As a result, the architecture combines a signal smoother and classification procedures into a single model. Validation results from rotor vibration data demonstrate that our model outperforms all other off-the-shelf feature-based models.

• Earthquakes and Structures
• /
• v.16 no.2
• /
• pp.165-175
• /
• 2019

Incremental dynamic analysis (IDA), as an accurate method to evaluate the parameters of structural performance levels, requires many non-linear time history analyses, using a set of ground motion records which are scaled to different intensity levels. Therefore, this method is very computationally demanding. In this study, a new method is presented to estimate the summarized (16%, 50%, and 84% fractiles) IDA curves of a first-mode dominated structure using discrete wavelet transform and bees optimization algorithm. This method reduces the number of required ground motion records for the prediction of the summarized IDA curves. At first, a subset of first list ground motion records is decomposed by means of discrete wavelet transform which have a low dispersion estimating the summarized IDA curves of equivalent SDOF system of the main structure. Then, the bees algorithm optimizes a series of factors for each level of detail coefficients in discrete wavelet transform. The applied factors change the frequency content of original ground motion records which the generated ground motions records can be utilized to reliably estimate the summarized IDA curves of the main structure. At the end, to evaluate the efficiency of the proposed method, the seismic behavior of a typical 3-story special steel moment frame, subjected to a set of twenty ground motion records is compared with this method.

• Journal of Intelligence and Information Systems
• /
• v.5 no.1
• /
• pp.103-123
• /
• 1999

Input filtering as a preprocessing method is so much crucial to get good performance in time series forecasting. There are a few preprocessing methods (i.e. ARMA outputs as time domain filters, and Fourier transform or wavelet transform as time-frequency domain filters) for handling time series. Specially, the time-frequency domain filters describe the fractal structure of financial markets better than the time domain filters due to theoretically additional frequency information. Therefore, we, first of all, try to describe and analyze specially some issues on the effectiveness of different filtering methods from viewpoint of the performance of a neural network based forecasting. And then we discuss about neural network model architecture issues, for example, what type of neural network learning architecture is selected for our time series forecasting, and what input size should be applied to a model. In this study an input selection problem is limited to a size selection of the lagged input variables. To solve this problem, we simulate on analyzing and comparing a few neural networks having different model architecture and also use an embedding dimension measure as chaotic time series analysis or nonlinear dynamic analysis to reduce the dimensionality (i.e. the size of time delayed input variables) of the models. Throughout our study, experiments for integration methods of joint time-frequency analysis and neural network techniques are applied to a case study of daily Korean won / U. S dollar exchange returns and finally we suggest an integration framework for future research from our experimental results.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.10a
• /
• pp.585-588
• /
• 2018

시계열 데이터의 특징을 추출하여 분석하는 과정에서 시게열 데이터가 가지는 고차원성은 차원의 저주(Course of Dimensionality)로 인해 데이터내의 유효한 정보를 찾는데 어려움을 만든다. 이러한 문제를 해결하기 위해 차원 축소 기법(dimensionality reduction)이 널리 사용되고 있지만, 축소 과정에서 발생하는 정보의 희석으로 인하여 시계열 데이터에 대한 군집화(clustering)등을 수행하는데 있어서 성능의 변화를 가져온다. 본 논문은 이러한 현상을 관찰하기 위해 이산 웨이블릿 변환(Discrete Wavelet Transform:DWT)과 오토 인코더(AutoEncoder)를 차원 축소 기법으로 활용하여 시계열 데이터의 차원을 압축 한 뒤, 압축된 데이터를 K-평균(K-means) 알고리즘에 적용하여 군집화의 효율성을 비교하였다. 성능 비교 결과, DWT는 압축된 차원수 그리고 오토인코더는 시계열 데이터에 대한 충분한 학습이 각각 보장된다면 좋은 군집화 성능을 보이는 것을 확인하였다.

• Proceedings of the KIPE Conference
• /
• 2014.07a
• /
• pp.522-523
• /
• 2014

Analysis of faults in induction motors has become a major field of research due to importance of loss and damage reduction and maximum online performance of motors. There are several methods to analyze the faults in an induction motor from conventional Fourier transform to modern decision-making neural networks. Considering detectability of fault among all methods, a new fault detection solution has been proposed; it is called as frequency-domain Discrete Wavelet Transform (FD-DWT). In this method, the stator current is decomposed through series of low- and high-pass filters and consequently, the fault characteristics are more visible, because additional components have been reduced. The objective of this paper is early detection of input voltage unbalance in induction motor using wavelet transform in frequency domain. Experimental results show the effectiveness of the proposed method in early detection of faults.

• The Korean Journal of Applied Statistics
• /
• v.25 no.6
• /
• pp.897-915
• /
• 2012

This paper examines the relationship between the stock and futures markets in terms of lead-lag relationship, correlation and the hedge ratio using wavelet analysis. The basic finding is that the relationship between the two markets significantly depends on the time-scale. First, there is a feedback relationship between the stock and futures markets in the long-run scale; however, weaker evidence is observed in shorter-run scales. Second, wavelet correlation between the two markets increases for a longer time scale. Third, the hedge ratio and the effectiveness of hedging strategies increase as the investment horizon gets longer. The results in this paper indicate that the stock and futures series are perfectly correlated in the long run and are tied together over long horizons.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.26 no.5
• /
• pp.513-518
• /
• 2008

The research of global-scale mass redistribution and it changed by Earth gravity filed variation observations, including Earth's oblateness $J_2$(also called low degree spherical harmonic coefficient $C_{20}$), is in continuous progress. Recently, the comparative analysis of geodetic observation SLR can be made by the development of GRACE and other time-variable gravity measurements. In this study, $C_{20}$ time series changes in the value of comparative analysis was got by GRACE monthly Gravity filed model (CSR RL04) for the period April 2002 to May 2008. And comparative analysis the harmonic coefficients of $C_{20}$ was obtained from SLR observations. Signal analysis for two time-series data was made by wavelet transform, CWT(continuous wavelet transform), XWT(cross wavelet transform) and WTC(wavelet coherence) methods. The results indicate that GRACE and SLR values for $C_{20}$ had both decreasing trend, as well as SLR data represent the annual frequencies, and GRACE was semiannual variations. In addition, the results of GRACE and SLR had a strong correlation with the XWT and WTC in an annual cycle.