Irregular Time Series Research Articles

주택용 계시별 요금제 (ToU) 대응을 위한 스마트플러그의 불규칙 전력 시계열 데이터 분류 AI 개발

본 연구에서는 우리나라 차후 도입 예정인 가정용 ToU(Time of Use) 요금 체계에서 스마트플러그에 체결된 전자기기의 요금 절감 가능 여부를 판별하는 AI를 개발하였다. 이를 위해, 스마트플러그에 체결된 전자기기의 사용 패턴을 파악하고 분석하는데 필요한 전력 사용 데이터를 약 3분 간격으로 11개월 동안 167개의 스마트플러그로 수집하였다. 이 과정에서 시계열 데이터에 불규칙성이 야기 되었고, 이를 해결하기 위하여 소모 전력의 백분위수를 추출하여 학습 데이터를 확보하였다. 완전 연결 인공신경망을 통해 데이터를 학습하였고 AI의 이진분류 정확도는 91.2%로 나타났다. 최종적으로 스마트플러그를 ToU 적용 절감 가능 여부에 따라 두 그룹으로 나누어 평균단가의 분포(평균 및 분산) 비교를 통해 인공지능의 적합성 여부를 평가하였다.

Extracting the Oscillatory Component and Defining a Mean Amplitude of Thermokinetic Oscillations in the H/Pd System

Self-sustained oscillations may help in the research of catalytic reactions. But analysis of the resulting nonlinear time series may be complicated, not least due to the lack of a definite baseline. A mathematical method to transform such irregular time series into a more useful form that oscillates around the zero line has been proposed, which preserves both the frequency and the key topological aspects of the originals in the process. To this end, the mean value theorem for integrals has been applied in constructing a base curve for nonequilibrium, periodic thermokinetic oscillations, q(t), recorded microcalorimetrically in two experiments, respectively, with oscillatory sorptions of H2 and D2 in Pd. Once the mean values are so calculated for each period of q(t), they are subjected to cubic spline interpolation to form a new nonoscillatory curve, h(t). The latter can be used as an essential baseline for the oscillatory component of the original thermokinetic time series. Crucially, the areas under both q(t) and h(t) are supposed to be strictly identical. This being confirmed, the pointwise subtraction q(t) – h(t) yields another oscillatory time series g(t), considered to be the oscillatory component extracted from the original thermokinetic data. Subsequently, the method has been applied to a further group of nine microcalorimetric time series q(t) representing thermokinetic oscillations in the H/Pd system. Using the so-constructed g(t) curves (i.e., extracted oscillatory components), a new parameter, the mean amplitude, has been defined and used for correlating the observed gradation of intensities (i.e., amplitudes) of thermokinetic oscillations with a range of experimental conditions applied. The mean amplitude turns out to be a linear function of the first ionization potential of noble gases, admixed intentionally to H2 before its contacting Pd.

Reconstruction of Sentinel-2 Image Time Series Using Google Earth Engine

Sentinel-2 NDVI and surface reflectance time series have been widely used in various geoscience research, but the data is deteriorated or missing due to the cloud contamination, so it is necessary to reconstruct the Sentinel-2 NDVI and surface reflectance time series. At present, there are few studies on reconstructing the Sentinel-2 NDVI or surface reflectance time series, and these existing reconstruction methods have some shortcomings. We proposed a new method to reconstruct the Sentinel-2 NDVI and surface reflectance time series using the penalized least-square regression based on discrete cosine transform (DCT-PLS) method. This method iteratively identifies cloud-contaminated NDVI over NDVI time series from the Sentinel-2 surface reflectance data by adjusting the weights. The NDVI and surface reflectance time series are then reconstructed from cloud-free NDVI and surface reflectance using the adjusted weights as constraints. We have made some improvements to the DCT-PLS method. First, the traditional discrete cosine transformation (DCT) in the DCT-PLS method is matrix generated from discrete and equally spaced data, we reconfigured the DCT formulas to adapt for irregular interval time series, and optimized the control parameters N and s according to the typical vegetation samples in China. Second, the DCT-PLS method was deployed in the Google Earth Engine (GEE) platform for the efficiency and convenience of data users. We used the DCT-PLS method to reconstruct the Sentinel-2 NDVI time series and surface reflectance time series in the blue, green, red, and near infrared (NIR) bands in typical vegetation samples and the Zhangjiakou and Hangzhou study area. We found that this method performed better than the SG filter method in reconstructing the NDVI time series, and can identify and reconstruct the contaminated NDVI as well as surface reflectance with low root mean square error (RMSE) and high coefficient of determination (R2). However, in cases of a long range of cloud contamination, or above water surface, it may be necessary to increase the control parameter s for a more stable performance. The GEE code is freely available online and the link is in the conclusions of this article, researchers are welcome to use this method to generate cloudless Sentinel-2 NDVI and surface reflectance time series with 10 m spatial resolution, which is convenient for landcover classification and many other types of research.

HTIDB: Hierarchical Time-Indexed Database for Efficient Storage and Access to Irregular Time-series Health Sensor Data.

With the enormous amount of data collected by unobtrusive sensors, the potential of utilizing these data and applying various multi-modal advanced analytics on them is numerous and promising. However, taking advantage of the ever-growing data requires high-performance data-handling systems to enable high data scalability and easy data accessibility. This paper demonstrates robust design, developments, and techniques of a hierarchical time-indexed database for decision support systems leveraging irregular and sporadic time series data from sensor systems, e.g., wearables or environmental. We propose a technique that leverages the flexibility of general purpose, high-scalability database systems, while integrating data analytics focused column stores that leverage hierarchical time indexing, compression, and dense raw numeric data storage. We have evaluated the performance characteristics and tradeoffs of each to understand the data access latencies and storage requirements, which are key elements for capacity planning for scalable systems.

Energy Consumption Forecasting in Korea Using Machine Learning Algorithms

In predicting energy consumption, classic econometric and statistical models are used to forecast energy consumption. These models may have limitations in an increasingly fast-changing energy market, which requires big data analysis of energy consumption patterns and relevant variables using complex mathematical tools. In current literature, there are minimal comparison studies reviewing machine learning algorithms to predict energy consumption in Korea. To bridge this gap, this paper compared three different machine learning algorithms, namely the Random Forest (RF) model, XGBoost (XGB) model, and Long Short-Term Memory (LSTM) model. These algorithms were applied in Period 1 (prior to the onset of the COVID-19 pandemic) and Period 2 (after the onset of the COVID-19 pandemic). Period 1 was characterized by an upward trend in energy consumption, while Period 2 showed a reduction in energy consumption. LSTM performed best in its prediction power specifically in Period 1, and RF outperformed the other models in Period 2. Findings, therefore, suggested the applicability of machine learning to forecast energy consumption and also demonstrated that traditional econometric approaches may outperform machine learning when there is less unknown irregularity in the time series, but machine learning can work better with unexpected irregular time series data.

LANDSLIDE INVENTORY MAPPING FROM LANDSAT-8 NDVI TIME SERIES USING ADAPTIVE LANDSLIDE INTERVAL DETECTION

Abstract. A landslide inventory map is one of the essential sources of geospatial information for land resource management. This study proposes an interval-based landslide detection strategy using satellite images' time series vegetation index. Landslide trends to be abruptly changed with the landscape can be clearly detected using the time series vegetation index. The proposed adaptive landslide interval detection (LID) method is a two stage algorithm. The first stage extracts local extremes and divides the time series to obtain discriminative intervals. Then, a predefined threshold is applied to determine whether a landslide occurs in each interval. The experiment compares the proposed scheme and the traditional time series forest (TSF) algorithm on a pre-labeled dataset. In the comparison of the results obtained from TSF and LID, the validation results show that the proposed LID has better discriminative ability in real scenarios, and the landslide detection rate for large areas reaches 85%. But the TSF can only provide good results on well-defined and unmixing datasets. In addition, LID does not require a large number of training datasets and can be applied to irregular time series with various lengths. In summary, this study demonstrated that the LID global pattern concerned splitting strategy is more effective than TSF random interval segmentation.

Identification of ship motion performance of coupled heave and pitch using experimental irregular time-series data

Identification of ship motion performance of coupled heave and pitch using experimental irregular time-series data

PPP-derived tropospheric ZWD augmentation from local CORS network tested on bridge monitoring points

This paper developed the tropospheric correction augmented PPP for the purpose of bridge displacement monitoring. ZWD is estimated from undifferenced observations collected by UK CORS, and further applied to augment PPP client for displacement at the bridge surveying points. Coordinate time series obtained by GPS DD are treated as the true bridge displacement, which is used to assess the standard and CORS ZWD augmented PPP solutions obtained by varying cutoff angles. The experimental results reveal that the displacement precision of standard PPP severely depends on the satellite geometry. Meanwhile, the CORS ZWD augmented PPP results are less affected. The RMSE of improved PPP time series in the vertical component drops by 12%, 49% and 66% for the 5°, 10° and 15° cutoff angle cases, respectively. CORS ZWD augmented PPP can improve the monitoring precision and eliminate irregular displacement time series due to poor satellite geometry.

Modeling of Soft Pneumatic Actuators with Different Orientation Angles Using Echo State Networks for Irregular Time Series Data.

Modeling of soft robotics systems proves to be an extremely difficult task, due to the large deformation of the soft materials used to make such robots. Reliable and accurate models are necessary for the control task of these soft robots. In this paper, a data-driven approach using machine learning is presented to model the kinematics of Soft Pneumatic Actuators (SPAs). An Echo State Network (ESN) architecture is used to predict the SPA’s tip position in 3 axes. Initially, data from actual 3D printed SPAs is obtained to build a training dataset for the network. Irregular-intervals pressure inputs are used to drive the SPA in different actuation sequences. The network is then iteratively trained and optimized. The demonstrated method is shown to successfully model the complex non-linear behavior of the SPA, using only the control input without any feedback sensory data as additional input to the network. In addition, the ability of the network to estimate the kinematics of SPAs with different orientation angles is achieved. The ESN is compared to a Long Short-Term Memory (LSTM) network that is trained on the interpolated experimental data. Both networks are then tested on Finite Element Analysis (FEA) data for other angle SPAs not included in the training data. This methodology could offer a general approach to modeling SPAs with varying design parameters.

An Interpretable Early Dynamic Sequential Predictor for Sepsis-Induced Coagulopathy Progression in the Real-World Using Machine Learning.

Sepsis-associated coagulation dysfunction greatly increases the mortality of sepsis. Irregular clinical time-series data remains a major challenge for AI medical applications. To early detect and manage sepsis-induced coagulopathy (SIC) and sepsis-associated disseminated intravascular coagulation (DIC), we developed an interpretable real-time sequential warning model toward real-world irregular data. Eight machine learning models including novel algorithms were devised to detect SIC and sepsis-associated DIC 8n (1 ≤ n ≤ 6) hours prior to its onset. Models were developed on Xi'an Jiaotong University Medical College (XJTUMC) and verified on Beth Israel Deaconess Medical Center (BIDMC). A total of 12,154 SIC and 7,878 International Society on Thrombosis and Haemostasis (ISTH) overt-DIC labels were annotated according to the SIC and ISTH overt-DIC scoring systems in train set. The area under the receiver operating characteristic curve (AUROC) were used as model evaluation metrics. The eXtreme Gradient Boosting (XGBoost) model can predict SIC and sepsis-associated DIC events up to 48 h earlier with an AUROC of 0.929 and 0.910, respectively, and even reached 0.973 and 0.955 at 8 h earlier, achieving the highest performance to date. The novel ODE-RNN model achieved continuous prediction at arbitrary time points, and with an AUROC of 0.962 and 0.936 for SIC and DIC predicted 8 h earlier, respectively. In conclusion, our model can predict the sepsis-associated SIC and DIC onset up to 48 h in advance, which helps maximize the time window for early management by physicians.

SAEP: A Surrounding-Aware Individual Emotion Prediction Model Combined with T-LSTM and Memory Attention Mechanism

The future emotion prediction of users on social media has been attracting increasing attention from academics. Previous studies on predicting future emotion have focused on the characteristics of individuals’ emotion changes; however, the role of the individual’s neighbors has not yet been thoroughly researched. To fill this gap, a surrounding-aware individual emotion prediction model (SAEP) based on a deep encoder–decoder architecture is proposed to predict individuals’ future emotions. In particular, two memory-based attention networks are constructed: The time-evolving attention network and the surrounding attention network to extract the features of the emotional changes of users and neighbors, respectively. Then, these features are incorporated into the emotion prediction task. In addition, a novel variant LSTM is introduced as the encoder of the proposed model, which can effectively extract complex patterns of users’ emotional changes from irregular time series. Extensive experimental results show that the proposed approach outperforms five alternative methods. The SAEP approach has improved by approximately 4.21–14.84% micro F1 on a dataset built from Twitter and 7.30–13.41% on a dataset built from Microblog. Further analyses validate the effectiveness of the proposed time-evolving context and surrounding context, as well as the factors that may affect the prediction results.

Comparing estimation techniques for temporal scaling in palaeoclimate time series

Abstract. Characterizing the variability across timescales is important for understanding the underlying dynamics of the Earth system. It remains challenging to do so from palaeoclimate archives since they are more often than not irregular, and traditional methods for producing timescale-dependent estimates of variability, such as the classical periodogram and the multitaper spectrum, generally require regular time sampling. We have compared those traditional methods using interpolation with interpolation-free methods, namely the Lomb–Scargle periodogram and the first-order Haar structure function. The ability of those methods to produce timescale-dependent estimates of variability when applied to irregular data was evaluated in a comparative framework, using surrogate palaeo-proxy data generated with realistic sampling. The metric we chose to compare them is the scaling exponent, i.e. the linear slope in log-transformed coordinates, since it summarizes the behaviour of the variability across timescales. We found that, for scaling estimates in irregular time series, the interpolation-free methods are to be preferred over the methods requiring interpolation as they allow for the utilization of the information from shorter timescales which are particularly affected by the irregularity. In addition, our results suggest that the Haar structure function is the safer choice of interpolation-free method since the Lomb–Scargle periodogram is unreliable when the underlying process generating the time series is not stationary. Given that we cannot know a priori what kind of scaling behaviour is contained in a palaeoclimate time series, and that it is also possible that this changes as a function of timescale, it is a desirable characteristic for the method to handle both stationary and non-stationary cases alike.

Performance Evaluation of Effective Drought Prediction Using Machine Learning

There has been much research recently to improve the prediction of drought, but the frequency and pattern of drought displays an irregular time series that limits its predictability, making it difficult to predict with only a single model, and high-level predictions cannot be made even when many models are applied. Therefore, many studies have been conducted to improve predictions by using explanatory variables such as precipitation, temperature, sunshine duration, and air volume as input data. The purpose of this study is to devise a method for predicting drought using the Standard Precipitation Evaporation Index (SPEI), which represents a complex and difficult time series drought index using climate data for weather phenomena. The Standard Precipitation Evaporation Index is a method of calculating the cumulative precipitation by excluding the cumulative evaporation amount from the cumulative precipitation using precipitation and evapotranspiration data, and the evaporation amount is calculated using the monthly heat index method. The Meteorological Agency evaluated meteorological drought using SPI6, which is a 6-month cumulative precipitation standard, and applied it to machine learning based on monthly data and daily data SPEI6 in this study. As a result, ANN monthly data R2 was 0.488 in Andong and 0.533 in Mungyeong, Gumi 0.594, SVR 0.452, 0.496, 0.564, RF 0.355, 0.467, 0.524, and the daily data are ANN 0.923, 0.919, 0.915, SVR 0.925, 0.923, 0.896, RF 0.915, 0.915, 0.797, and the daily data SPEI at all points. It was confirmed that high prediction was obtained when machine learning was applied to these methods.

Incorporating real-world evidence into the development of patient blood glucose prediction algorithms for the ICU

ObjectiveGlycemic control is an important component of critical care. We present a data-driven method for predicting intensive care unit (ICU) patient response to glycemic control protocols while accounting for patient heterogeneity and variations in care.Materials and MethodsUsing electronic medical records (EMRs) of 18 961 ICU admissions from the MIMIC-III dataset, including 318 574 blood glucose measurements, we train and validate a gradient boosted tree machine learning (ML) algorithm to forecast patient blood glucose and a 95% prediction interval at 2-hour intervals. The model uses as inputs irregular multivariate time series data relating to recent in-patient medical history and glycemic control, including previous blood glucose, nutrition, and insulin dosing.ResultsOur forecasting model using routinely collected EMRs achieves performance comparable to previous models developed in planned research studies using continuous blood glucose monitoring. Model error, expressed as mean absolute percentage error is 16.5%–16.8%, with Clarke error grid analysis demonstrating that 97% of predictions would be clinically acceptable. The 95% prediction intervals achieve near intended coverage at 93%–94%.DiscussionML algorithms built on observational data sources, such as EMRs, present a promising approach for personalization and automation of glycemic control in critical care. Future research may benefit from applying a combination of methodologies and data sources to develop robust methodologies that account for the variations seen in ICU patients and difficultly in detecting the extremes of observed blood glucose values.ConclusionWe demonstrate that EMRs can be used to train ML algorithms that may be suitable for incorporation into ICU decision support systems.

Short-term wind power prediction based on EEMD–LASSO–QRNN model

With the increasing utilization of wind generation in power system, the improvement of wind power forecasting precision is attached vital importance. Owing to the stochastic and intermittent nature of wind power, the conventional methods no longer ensure sufficient accuracy of wind power prediction in majority of scenarios. Motivated by recent advancements of ensemble methods based on decomposition technologies, a novel ensemble method based on ensemble empirical mode decomposition (EEMD) and least absolute shrinkage and selection operator–quantile regression neural network (LASSO–QRNN) model for forecasting wind power is proposed in this paper. The model is an ingenious integration of data preprocessing technology, feature selection method, prediction model and data post-processing technology. Thereinto, EEMD is exploited to convert intricate and irregular wind power time series into a collection of subseries relatively easy to analyze; LASSO regression is combined with QRNN model to realize the filtering of important variables and provide more comprehensive and robust prediction results; the KDE method reprocesses the prediction results, greatly improves the prediction accuracy and effectively quantifies the uncertainty of the forecasting process. The suggested model and several benchmark models have been implemented on six wind power datasets, two are gathered from a wind farm in Spain and four from a competition to demonstrate the superiorities of the model proposed in this paper. The compared results reveal that the proposed method has adequate capacity to enhance the performance of wind power forecasting, measure and reduce the uncertainty of prediction process.

A review of irregular time series data handling with gated recurrent neural networks

Irregular time series data is becoming increasingly prevalent with the growth of multi-sensor systems as well as the continued use of unstructured manual data recording mechanisms. Irregular data and the resulting missing values severely limit the data's ability to be analysed and modelled for classification and forecasting tasks. Often, conventional methods used for handling time series data introduce bias and make strong assumptions on the underlying data generation process, which can lead to poor model predictions. Traditional machine learning and deep learning methods, although at the forefront of data modelling, are at best compromised by irregular time series data sets and fail to model the temporal irregularity of incomplete time series. Gated recurrent neural networks (RNN), such as LSTM and GRU, have had outstanding success in sequential modelling, and have been applied in many application fields, including natural language processing. These models have become an obvious choice for time series modelling and a promising tool for handling irregular time series data. RNNs have a unique ability to be adapted to make effective use of missing value patterns, time intervals and complex temporal dependencies in irregular univariate and multivariate time series data. In this paper, we provide a systematic review of recent studies in which gated recurrent neural networks have been successfully applied to irregular time series data for prediction tasks within several fields, including medical, human activity recognition, traffic monitoring and environmental monitoring. The review highlights the two common approaches for handling irregular time series data: missing value imputation at the data pre-processing stage and modification of algorithms to directly handle missing values in the learning process. Reviewed models are confined to those that can address issues with irregular time series data and does not cover the broader range of models that deal more generally with sequences and regular time series. This paper aims to present the most effective techniques emerging within this branch of research as well as to identify remaining challenges, so that researchers may build upon this platform of work towards further novel techniques for handling irregular time series data.

Arm order recognition in multi-armed bandit problem with laser chaos time series

By exploiting ultrafast and irregular time series generated by lasers with delayed feedback, we have previously demonstrated a scalable algorithm to solve multi-armed bandit (MAB) problems utilizing the time-division multiplexing of laser chaos time series. Although the algorithm detects the arm with the highest reward expectation, the correct recognition of the order of arms in terms of reward expectations is not achievable. Here, we present an algorithm where the degree of exploration is adaptively controlled based on confidence intervals that represent the estimation accuracy of reward expectations. We have demonstrated numerically that our approach did improve arm order recognition accuracy significantly, along with reduced dependence on reward environments, and the total reward is almost maintained compared with conventional MAB methods. This study applies to sectors where the order information is critical, such as efficient allocation of resources in information and communications technology.

Inference of population genetic parameters from an irregular time series of seasonal influenza virus sequences.

Basic summary statistics that quantify the population genetic structure of influenza virus are important for understanding and inferring the evolutionary and epidemiological processes. However, the sampling dates of global virus sequences in the last several decades are scattered nonuniformly throughout the calendar. Such temporal structure of samples and the small effective size of viral population hampers the use of conventional methods to calculate summary statistics. Here, we define statistics that overcome this problem by correcting for the sampling-time difference in quantifying a pairwise sequence difference. A simple linear regression method jointly estimates the mutation rate and the level of sequence polymorphism, thus providing an estimate of the effective population size. It also leads to the definition of Wright's FST for arbitrary time-series data. Furthermore, as an alternative to Tajima's D statistic or the site-frequency spectrum, a mismatch distribution corrected for sampling-time differences can be obtained and compared between actual and simulated data. Application of these methods to seasonal influenza A/H3N2 viruses sampled between 1980 and 2017 and sequences simulated under the model of recurrent positive selection with metapopulation dynamics allowed us to estimate the synonymous mutation rate and find parameter values for selection and demographic structure that fit the observation. We found that the mutation rates of HA and PB1 segments before 2007 were particularly high and that including recurrent positive selection in our model was essential for the genealogical structure of the HA segment. Methods developed here can be generally applied to population genetic inferences using serially sampled genetic data.

Fault Classification of Nonlinear Small Sample Data through Feature Sub-Space Neighbor Vote

The fault classification of a small sample of high dimension is challenging, especially for a nonlinear and non-Gaussian manufacturing process. In this paper, a similarity-based feature selection and sub-space neighbor vote method is proposed to solve this problem. To capture the dynamics, nonlinearity, and non-Gaussianity in the irregular time series data, high order spectral features, and fractal dimension features are extracted, selected, and stacked in a regular matrix. To address the problem of a small sample, all labeled fault data are used for similarity decisions for a specific fault type. The distances between the new data and all fault types are calculated in their feature subspaces. The new data are classified to the nearest fault type by majority probability voting of the distances. Meanwhile, the selected features, from respective measured variables, indicate the cause of the fault. The proposed method is evaluated on a publicly available benchmark of a real semiconductor etching dataset. It is demonstrated that by using the high order spectral features and fractal dimensionality features, the proposed method can achieve more than 84% fault recognition accuracy. The resulting feature subspace can be used to match any new fault data to the fingerprint feature subspace of each fault type, and hence can pinpoint the root cause of a fault in a manufacturing process.

Analysis of Pollution Patterns in Regions of Kolkata

This article analyses various patterns in the pollution levels of [Formula: see text], [Formula: see text], [Formula: see text], and [Formula: see text] at various locations in Kolkata, India. Our analysis is based on a data set recorded by the Central Pollution Control Board of India and the West Bengal Pollution Control Board, comprising of about 15 years of irregular time series observations, due to the higher cost of precise and accurate measurements of pollution. We perform some exploratory analysis concerning the variations in trend, seasonal, and shift-specific seasonal levels. As well as a statistical model for forecasting pollution levels for two to three days in advance is also obtained, along with an analysis of the effects of festivities such as Kali Puja and Diwali on the level of the pollutants across different locations.