• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.35 no.2
• /
• pp.106-112
• /
• 2012

Real-life time series characteristic data has significant amount of non-stationary components, especially periodic components in nature. Extracting such components has required many ad-hoc techniques with external parameters set by users in case-by-case manner. In our study, we evaluate whether Hilbert-Huang Transform, a new tool of time-series analysis can be used for effective analysis of such data. It is divided into two points : 1) how effective it is in finding periodic components, 2) whether we can use its results directly in detecting values outside control limits, for which a traditional method such as ARIMA had been used. We use glass furnace temperature data to illustrate the method.

• Journal of Intelligence and Information Systems
• /
• v.1 no.1
• /
• pp.135-148
• /
• 1995

Neural Networks have been advocated as an alternative to statistical forecasting methods. However, the empirical evidences are not consistent. In the present experiments, multi-layered perceptron (MLP) are adopted as approximator to the time series generating processes. To prevent the MLP from being overfitted to the given time series, the information obtained from ARMA modeling is used to determine the architecture of MLP. The proposed approach was tested empirically using the subsamples of the 111 time series used in the first Markridakis Competition. The forecasting results were analyzed to find out the factors that affect the performance of MLP. The experimental results show that the proposed approach outperforms ARMA models in terms of fitting and forecasting accuracy. In addition, it is found that the use of deseasonalized data improves the forecasting accuracy of MLP.

• International Journal of Internet, Broadcasting and Communication
• /
• v.16 no.3
• /
• pp.321-328
• /
• 2024

This paper designs a disease prediction algorithm to diagnose migraine among the types of diseases in advance by learning algorithms using machine learning-based time series analysis. This study utilizes patient data statistics, such as electroencephalogram activity, to design a prediction algorithm to determine the onset signals of migraine symptoms, so that patients can efficiently predict and manage their disease. The results of the study evaluate how accurate the proposed prediction algorithm is in predicting migraine and how quickly it can predict the onset of migraine for disease prevention purposes. In this paper, a machine learning algorithm is used to analyze time series of data indicators used for migraine identification. We designed an algorithm that can efficiently predict and manage patients' diseases by quickly determining the onset signaling symptoms of disease development using existing patient data as input. The experimental results show that the proposed prediction algorithm can accurately predict the occurrence of migraine using machine learning algorithms.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.377-379
• /
• 2023

Concept drift detection on data stream is the major issue to maintain the performance of the machine learning model. Since the online stream is to be a function of time, the classical statistic methods are hard to apply. In particular case of seasonal time series, a novel window strategy with Fourier analysis however, gives a chance to adapt the classical methods on the series. We explore the KS-test for an adaptation of the periodic time series and show that this strategy handles a complicate time series as an ordinary tabular dataset. We verify that the detection with the strategy takes the second place in time delay and shows the best performance in false alarm rate and detection accuracy comparing to that of arbitrary window sizes.

• Journal of IKEEE
• /
• v.24 no.4
• /
• pp.1011-1016
• /
• 2020

We propose a framework called Stacked Gated Recurrent Unit - Infrequent Residual Analysis (SG-IRA) that detects anomalies in time-series data that can be trained on streams of raw sensor data without any pre-labeled dataset. To enable such unsupervised learning, SG-IRA includes an estimation model that uses a stacked Gated Recurrent Unit (GRU) structure and an analysis method that detects anomalies based on the difference between the estimated value and the actual measurement (residual). SG-IRA's residual analysis method dynamically adapts the detection threshold from the population using frequency analysis, unlike the baseline model that relies on a constant threshold. In this paper, SG-IRA is evaluated using the industrial control systems (ICS) datasets. SG-IRA improves the detection performance (F1 score) by 5.9% compared to the baseline model.

• Journal of Korea Water Resources Association
• /
• v.48 no.12
• /
• pp.981-993
• /
• 2015

The analysis of hydrologic time series data is crucial for the effective management of water resources. Therefore, it has been widely used for the long-term forecasting of hydrologic variables. In tradition, time series analysis has been used to predict a time series without considering exogenous variables. However, many studies using decomposition have been widely carried out with the assumption that one data series could be mixed with several frequent factors. In this study, the empirical mode decomposition method was performed for decomposing a hydrologic time series data into several components, and each component was applied to the time series models, autoregressive moving average (ARMA). After constructing the time series models, the forecasting values are added to compare the results with traditional time series model. Finally, the forecasted estimates from ARMA model with empirical mode decomposition method showed better performance than sole traditional ARMA model indicated from comparing the root mean square errors of the two methods.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.46 no.3
• /
• pp.123-138
• /
• 2023

The Fourth Industrial Revolution and sensor technology have led to increased utilization of sensor data. In our modern society, data complexity is rising, and the extraction of valuable information has become crucial with the rapid changes in information technology (IT). Recurrent neural networks (RNN) and long short-term memory (LSTM) models have shown remarkable performance in natural language processing (NLP) and time series prediction. Consequently, there is a strong expectation that models excelling in NLP will also excel in time series prediction. However, current research on Transformer models for time series prediction remains limited. Traditional RNN and LSTM models have demonstrated superior performance compared to Transformers in big data analysis. Nevertheless, with continuous advancements in Transformer models, such as GPT-2 (Generative Pre-trained Transformer 2) and ProphetNet, they have gained attention in the field of time series prediction. This study aims to evaluate the classification performance and interval prediction of remaining useful life (RUL) using an advanced Transformer model. The performance of each model will be utilized to establish a health index (HI) for cutting blades, enabling real-time monitoring of machine health. The results are expected to provide valuable insights for machine monitoring, evaluation, and management, confirming the effectiveness of advanced Transformer models in time series analysis when applied in industrial settings.

• Proceedings of the KIEE Conference
• /
• 1984.07a
• /
• pp.68-70
• /
• 1984

• Journal of Preventive Medicine and Public Health
• /
• v.46 no.1
• /
• pp.10-18
• /
• 2013

Objectives: We investigated the association between particulate matter less than $10{\mu}m$ in aerodynamic diameter ($PM_{10}$) exposure and non-accidental mortality in Asian populations by meta-analysis, using both time-series and case-crossover analysis. Methods: Among the 819 published studies searched from PubMed and EMBASE using key words related to $PM_{10}$ exposure and non-accidental mortality in Asian countries, 8 time-series and 4 case-crossover studies were selected for meta-analysis after exclusion by selection criteria. We obtained the relative risk (RR) and 95% confidence intervals (CI) of non-accidental mortality per $10{\mu}g/m^3$ increase of daily $PM_{10}$ from each study. We used Q statistics to test the heterogeneity of the results among the different studies and evaluated for publication bias using Begg funnel plot and Egger test. Results: Testing for heterogeneity showed significance (p<0.001); thus, we applied a random-effects model. RR (95% CI) per $10{\mu}g/m^3$ increase of daily $PM_{10}$ for both the time-series and case-crossover studies combined, time-series studies relative risk only, and case-crossover studies only, were 1.0047 (1.0033 to 1.0062), 1.0057 (1.0029 to 1.0086), and 1.0027 (1.0010 to 1.0043), respectively. The non-significant Egger test suggested that this analysis was not likely to have a publication bias. Conclusions: We found a significant positive association between $PM_{10}$ exposure and non-accidental mortality among Asian populations. Continued investigations are encouraged to contribute to the health impact assessment and public health management of air pollution in Asian countries.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.54 no.7
• /
• pp.462-467
• /
• 2005

In this paper, we analyzed normal gait acceleration signal by time series analysis methods. Accelerations were measured during walking using a biaxial accelerometer. Acceleration data were acquired from normal subjects(23 men and one woman) walking on a level corridor of 20m in length with three different walking speeds. Acceleration signals were measured at a sampling frequency of 60Hz from a biaxial accelerometer mounted between L3 and L4 intervertebral area. Each step signal was analyzed using Box-Jenkins method. Most of the differenced normal step signals were modeled to AR(3) and the model didn't show difference for model's orders and coefficients with walking speed. But, tile model showed difference with acceleration signal direction - vertical and lateral. The above results suggested the proposed model could be applied to unit analysis.