• The Korean Journal of Applied Statistics
• /
• v.32 no.6
• /
• pp.851-865
• /
• 2019

Discovering various features from one time series is complicated. In this paper, we introduce a multi aggregation prediction algorithm (MAPA) that uses the concepts of temporal aggregation and combining forecasts to find multiple patterns from one time series and increase forecasting accuracy. Temporal aggregation produces multiple time series and each series has separate properties. We use exponential smoothing methods in the next step to extract various features of time series components in order to forecast time series components for each series. In the final step, we blend predictions of the same kind of components and forecast the target series by the summation of blended predictions. As an empirical example, we forecast traffic accident counts using MAPA and observe that MAPA performance is superior to conventional methods.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.631-634
• /
• 2004

In this paper, the Bayesian recurrent neural network (BRNN) is proposed to predict time series data. Among the various traditional prediction methodologies, a neural network method is considered to be more effective in case of non-linear and non-stationary time series data. A neural network predictor requests proper learning strategy to adjust the network weights, and one need to prepare for non-linear and non-stationary evolution of network weights. The Bayesian neural network in this paper estimates not the single set of weights but the probability distributions of weights. In other words, we sets the weight vector as a state vector of state space method, and estimates its probability distributions in accordance with the Bayesian inference. This approach makes it possible to obtain more exact estimation of the weights. Moreover, in the aspect of network architecture, it is known that the recurrent feedback structure is superior to the feedforward structure for the problem of time series prediction. Therefore, the recurrent network with Bayesian inference, what we call BRNN, is expected to show higher performance than the normal neural network. To verify the performance of the proposed method, the time series data are numerically generated and a neural network predictor is applied on it. As a result, BRNN is proved to show better prediction result than common feedforward Bayesian neural network.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2004.10a
• /
• pp.426-429
• /
• 2004

To reconstruct the 3-D information of a irregular object, this paper proposes a new method applying the composition focus measure to time-series image. A focus measure function is carefully selected because a focus measure is apt to be affected by the working environment and the characteristics of an object. The proposed focus measure function combines the variance measure which is robust to noise and the Laplacian measure which, regardless of an object shape, has a good performance in calculating the focus measure. And the time-series image, which considers the object shape, is proposed in order to efficiently applying the interesting window. This method, first, divides the image frame by the window. Second, the composition focus measure function be applied to the windows, and the time-series image is constructed. Finally, the 3-D information of an object is reconstructed from the time-series images considering the object shape. The experimental results have shown that the proposed method is suitable algorithm to 3-D reconstruction of an irregular object.

• International Journal of Advanced Culture Technology
• /
• v.8 no.3
• /
• pp.216-222
• /
• 2020

The purpose of this study was to predict the performance of female weightlifters using time series analysis. Based on this purpose, a time series analysis was used to calculate the performance prediction model for women(58kg) among the domestic women weightlifters who participated in the Olympics. As a result of creating time series data based on 10 years of record and then evaluating the sequential charts of each athlete group, the female athletes' records did not show any seasonality or difference. In addition, after examining the independence of the data through the creation of a time series model, it was shown that the models produced conformed to the criteria for compliance and that there was no difference in the data, but there was a trend. Accordingly, Holt linear trend analysis of the exponential smoothing model was applied. As a result of deriving the prediction model of the athletes through this process, it was found that the women (58kg) who participated in the Olympics continued to improve within the range of 166.11kg to 184.1kg.

• Proceedings of the Korean Society of Soil and Groundwater Environment Conference
• /
• 2003.09a
• /
• pp.254-257
• /
• 2003

Statistical analysis is performed to estimate the correlations between geological or geographical factor and groundwater inflow rates in the Seoul subway system. Correlation analysis shows that among several geological and geographical factors fractures and streams have most strong effects on inflow rate into tunnels. In particular, subway line 5∼8 are affected more by these factors than subway line 1∼4. Time series analysis is carried out to forecast groundwater inflow rate. Time series analysis is a useful empirical method for simulation and forecasts in case that physical model can not be applied to. The time series of groundwater inflow rates is calculated using the observation data. Transfer function-noise model is applied with the precipitation data as input variables. For time series analysis, statistical methods are performed to identify proper model and autoregressive-moving average models are applied to evaluation of inflow rate. Each model is identified to satisfy the lowest value of information criteria. Results show that the values by result equations are well fitted with the actual inflow rate values. The selected models could give a good explanation of inflow rates variation into subway tunnels.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.5 s.43
• /
• pp.69-74
• /
• 2006

The recent development of multimedia content-based retrieval technologies brings great attention of musician retrieval using features of a digital audio data among music information retrieval technologies. But the indexing techniques for music databases have not been studied completely. In this paper, we present a musician retrieval technique for audio features using the space split methods in the time-series pattern index file. We use features of audio to retrieve the musician and a time-series pattern index file to search the candidate musicians. Experimental results show that the time-series pattern index file using the rotational split method is efficient for musician retrievals in the time-series pattern files.

• Journal of Internet of Things and Convergence
• /
• v.4 no.1
• /
• pp.1-5
• /
• 2018

This paper is concerned with the modeling and identification of time series data corrupted by noise. As modeling techniques, nonsingleton fuzzy logic system (NFLS) is employed for the modeling of corrupted time series. Main characteristic of the NFLS is a fuzzy system whose inputs are modeled as fuzzy number. So the NFLS is especially useful in cases where the available training data or the input data to the fuzzy logic system are corrupted by noise. Simulation results of the Mackey-Glass time series data will be demonstrated to show the performance of the modeling methods. As a result, NFLS does a much better job of modeling noisy time series data than does a traditional Mamdani FLS.

• The KIPS Transactions:PartD
• /
• v.14D no.7
• /
• pp.727-732
• /
• 2007

Most of previous work on indexing and searching time series focused on the similarity matching and retrieval of one-attribute time series. However, multimedia databases such as music, video need to handle the similarity search in multi-attribute time series. The limitation of the current similarity models for multi-attribute sequences is that there is no consideration for attributes' sequences. The multi-attribute sequences are composed of several attributes' sequences. Since the users may want to find the similar patterns considering attributes's sequences, it is more appropriate to consider the similarity between two multi-attribute sequences in the viewpoint of attributes' sequences. In this paper, we propose the similarity search method based on attributes's sequences in multi-attribute time series databases. The proposed method can efficiently reduce the search space and guarantees no false dismissals. In addition, we give preliminary experimental results to show the effectiveness of the proposed method.

• The Korean Journal of Applied Statistics
• /
• v.28 no.1
• /
• pp.115-122
• /
• 2015

This article is concerned with count time series taking values in non-negative integers. Along with the first order mean of the count time series, conditional variance (volatility) has recently been paid attention to and therefore various integer-valued GARCH(generalized autoregressive conditional heteroscedasticity) models have been suggested in the last decade. We introduce diverse integer-valued GARCH(INGARCH, for short) processes to count time series and a real data application is illustrated as a case study. In addition, zero inflated INGARCH models are discussed to accommodate zero-inflated count time series.

• Journal of Information Processing Systems
• /
• v.15 no.3
• /
• pp.694-706
• /
• 2019

In this study, we applied the long short-term memory (LSTM) model to classify the cryptocurrency price time series. We collected historic cryptocurrency price time series data and preprocessed them in order to make them clean for use as train and target data. After such preprocessing, the price time series data were systematically encoded into the three-dimensional price tensor representing the past price changes of cryptocurrencies. We also presented our LSTM model structure as well as how to use such price tensor as input data of the LSTM model. In particular, a grid search-based k-fold cross-validation technique was applied to find the most suitable LSTM model parameters. Lastly, through the comparison of the f1-score values, our study showed that the LSTM model outperforms the gradient boosting model, a general machine learning model known to have relatively good prediction performance, for the time series classification of the cryptocurrency price trend. With the LSTM model, we got a performance improvement of about 7% compared to using the GB model.