• Proceedings of the KSRS Conference
• /
• 1999.11a
• /
• pp.429-432
• /
• 1999

Vegetation cover classification is examined based on a time series NOAA/AVHRR data. Time series data analysis methods including Fourier transform, Auto-Regressive (AR) model and temporal signature similarity matching are developed to extract phenological features of vegetation from a time series NDVI data from NOAA/AVHRR and to classify vegetation types. In the Fourier transform method, typical three spectral components expressing the phenological features of vegetation are selected for classification, and also in the AR model method AR coefficients are selected. In the temporal signature similarity matching method a new index evaluating the similarity of temporal pattern of the NDVI is introduced for classification.

• The Journal of the Korea Contents Association
• /
• v.17 no.1
• /
• pp.39-49
• /
• 2017

As the Internet technology advances, data on the web is increasing sharply. Many research study about incremental learning for classifying effectively in data increasing. Web document contains the time-series data such as published date. If we reflect time-series data to classification, it will be an effective classification. In this study, we analyze the time-series variation of the words. We propose an efficient classification through dividing the dataset based on the analysis of time-series information. For experiment, we corrected 1 million online news articles including time-series information. We divide the dataset and classify the dataset using SVM and $Na{\ddot{i}}ve$ Bayes. In each model, we show that classification performance is increasing. Through this study, we showed that reflecting time-series information can improve the classification performance.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.4
• /
• pp.1-11
• /
• 2023

In this paper, instead of using recurrent neural network, we compare a classification performance of time series imaging algorithms using convolution neural network. There are traditional algorithms that imaging time series data (e.g. GAF(Gramian Angular Field), MTF(Markov Transition Field), RP(Recurrence Plot)) in TSC(Time Series Classification) community. Furthermore, we compare STFT(Short Time Fourier Transform) algorithm that can acquire spectrogram that visualize feature of voice data. We experiment CNN's performance by adjusting hyper parameters of imaging algorithms. When evaluate with GunPoint dataset in UCR archive, STFT(Short-Time Fourier transform) has higher accuracy than other algorithms. GAF has 98~99% accuracy either, but there is a disadvantage that size of image is massive.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.55-67
• /
• 2022

Sensor data can provide fault diagnosis for equipment. However, the cause analysis for fault results of equipment is not often provided. In this study, we propose an explainable convolutional neural network framework for the sensor-based time series classification model. We used sensor-based time series dataset, acquired from vehicles equipped with sensors, and the Wafer dataset, acquired from manufacturing process. Moreover, we used Cycle Signal dataset, acquired from real world mechanical equipment, and for Data augmentation methods, scaling and jittering were used to train our deep learning models. In addition, our proposed classification models are convolutional neural network based models, FCN, 1D-CNN, and ResNet, to compare evaluations for each model. Our experimental results show that the ResNet provides promising results in the context of time series classification with accuracy and F1 Score reaching 95%, improved by 3% compared to the previous study. Furthermore, we propose XAI methods, Class Activation Map and Layer Visualization, to interpret the experiment result. XAI methods can visualize the time series interval that shows important factors for sensor data classification.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.43 no.2
• /
• pp.33-38
• /
• 2020

Anomaly detection of Machine Learning such as PCA anomaly detection and CNN image classification has been focused on cross-sectional data. In this paper, two approaches has been suggested to apply ML techniques for identifying the failure time of big time series data. PCA anomaly detection to identify time rows as normal or abnormal was suggested by converting subjects identification problem to time domain. CNN image classification was suggested to identify the failure time by re-structuring of time series data, which computed the correlation matrix of one minute data and converted to tiff image format. Also, LASSO, one of feature selection methods, was applied to select the most affecting variables which could identify the failure status. For the empirical study, time series data was collected in seconds from a power generator of 214 components for 25 minutes including 20 minutes before the failure time. The failure time was predicted and detected 9 minutes 17 seconds before the failure time by PCA anomaly detection, but was not detected by the combination of LASSO and PCA because the target variable was binary variable which was assigned on the base of the failure time. CNN image classification with the train data of 10 normal status image and 5 failure status images detected just one minute before.

• 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.

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.329-332
• /
• 2007

Annual vegetation growth patterns are determined by the intrinsic phenological characteristics of each land cover types. So, if typical growth patterns of each land cover types are well-estimated, and a NDVI time-series data of a certain area is compared to those estimated patterns, we can implement more advanced analyses such as a land surface-type classification or a land surface type change detection. In this study, we utilized Terra MODIS NDVI 250m data and compressed full annual NDVI time series data into several indices using the Harmonic Analysis of Time Series(HANTS) algorithm which extracts the most significant frequencies expected to be presented in the original NDVI time-series data. Then, we found these frequencies patterns, described by amplitude and phase data, were significantly different from each other according to vegetation types and these could be used for land cover classification. However, in spite of the capabilities of the HANTS algorithm for detecting and interpolating cloud-contaminated NDVI values, some distorted NDVI pixels of June, July and August, as well as the long rainy season in Korea, are not properly corrected. In particular, in the case of two or three successive NDVI time-series data, which are severely affected by clouds, the HANTS algorithm outputted wrong results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.6
• /
• pp.1530-1544
• /
• 2023

Data monitoring is an important foundation of modern science. In most cases, the monitoring data is time-series data, which has high application value. The deep learning algorithm has a strong nonlinear fitting capability, which enables the recognition of time series by capturing anomalous information in time series. At present, the research of time series recognition based on deep learning is especially important for data monitoring. Deep learning algorithms require a large amount of data for training. However, abnormal sample is a small sample in time series, which means the number of abnormal time series can seriously affect the accuracy of recognition algorithm because of class imbalance. In order to increase the number of abnormal sample, a data augmentation method called GANBATS (GAN-based Bi-LSTM and Attention for Time Series) is proposed. In GANBATS, Bi-LSTM is introduced to extract the timing features and then transfer features to the generator network of GANBATS.GANBATS also modifies the discriminator network by adding an attention mechanism to achieve global attention for time series. At the end of discriminator, GANBATS is adding averagepooling layer, which merges temporal features to boost the operational efficiency. In this paper, four time series datasets and five data augmentation algorithms are used for comparison experiments. The generated data are measured by PRD(Percent Root Mean Square Difference) and DTW(Dynamic Time Warping). The experimental results show that GANBATS reduces up to 26.22 in PRD metric and 9.45 in DTW metric. In addition, this paper uses different algorithms to reconstruct the datasets and compare them by classification accuracy. The classification accuracy is improved by 6.44%-12.96% on four time series datasets.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.13 no.3
• /
• pp.186-199
• /
• 2013

We pose pattern classification as a density estimation problem where we consider mixtures of generative models under partially labeled data setups. Unlike traditional approaches that estimate density everywhere in data space, we focus on the density along the decision boundary that can yield more discriminative models with superior classification performance. We extend our earlier work on the recursive estimation method for discriminative mixture models to semi-supervised learning setups where some of the data points lack class labels. Our model exploits the mixture structure in the functional gradient framework: it searches for the base mixture component model in a greedy fashion, maximizing the conditional class likelihoods for the labeled data and at the same time minimizing the uncertainty of class label prediction for unlabeled data points. The objective can be effectively imposed as individual mixture component learning on weighted data, hence our mixture learning typically becomes highly efficient for popular base generative models like Gaussians or hidden Markov models. Moreover, apart from the expectation-maximization algorithm, the proposed recursive estimation has several advantages including the lack of need for a pre-determined mixture order and robustness to the choice of initial parameters. We demonstrate the benefits of the proposed approach on a comprehensive set of evaluations consisting of diverse time-series classification problems in semi-supervised scenarios.

• The Korean Journal of Applied Statistics
• /
• v.31 no.6
• /
• pp.707-719
• /
• 2018

Various time series representation methods have been proposed for efficient time series clustering and classification. Lin et al. (DMKD, 15, 107-144, 2007) proposed a symbolic aggregate approximation (SAX) method based on symbolic representations after approximating the original time series using piecewise local mean. The performance of SAX therefore depends heavily on how well the piecewise local averages approximate original time series features. SAX equally divides the entire series into an arbitrary number of segments; however, it is not sufficient to capture key features from complex, large-scale time series data. Therefore, this paper considers data-adaptive local constant approximation of the time series using the unbalanced Haar wavelet transformation. The proposed method is shown to outperforms SAX in many real-world data applications.