Related Time Series Research Articles

Wind Power Ramp Event Forecasting Based on Feature Extraction and Deep Learning

In order to improve the accuracy of wind power ramp forecasting and reduce the threat of ramps to the safe operation of power systems, a wind power ramp event forecast model based on feature extraction and deep learning is proposed in this work. Firstly, the Optimized Swinging Door Algorithm (OpSDA) is introduced to detect wind power ramp events, and the extraction results of ramp features, such as the ramp rate, are obtained. Then, a ramp forecast model based on a deep learning network is established. The historical wind power and its ramp features are used as the input of the forecast model, thereby strengthening the model’s learning for ramp features and preventing ramp features from being submerged in the complex wind power signal. A Convolutional Neural Network (CNN) is adopted to extract features from model inputs to obtain the coupling relationship between wind power and ramp features, and Long Short-Term Memory (LSTM) is utilized to learn the time-series relationship of the data. The forecast wind power is used as the output of the model, based on which the ramp forecast result is obtained after the ramp detection. Finally, the wind power data from the Elia website is used to verify the forecast performance of the proposed method for wind power ramp events.

Using the LSTM Network to Forecast the Demand for Electricity in Poland

The impact of environmental regulations introduced by the European Union is of key importance for electricity generation systems. The Polish fuel structure of electricity production is based on solid fuels. Moreover, the generating base is outdated and must gradually be withdrawn from the power system. In this context, Poland’s energy policy is undergoing a transformation as climate and environmental regulations are becoming increasingly stringent for the energy sector based on solid fuels (hard coal and lignite). However, the transformation process must be adapted to market demands, because the overriding goal is to ensure energy security by maintaining the continuity of energy supplies and an acceptable electricity price. This directly contributes to the development of the entire economy and the standard of living of the society, in accordance with the European Agreement establishing an association between the Republic of Poland and the European Communities and their Member States, signed on 16 December 1991, and the European Energy Charter, signed on 17 December 1991. Ensuring energy security is the most important goal of the energy policy. Therefore, energy companies must forecast the demand. The main goal of this article is to develop a mathematical model of electricity consumption by 2040 by all sectors of the economy: industry, transport, residential, commercial and public services, agriculture, forestry, and fishing. In order to achieve the intended goal, a model was developed by using Long Short-Term Memory (LSTM) artificial neural networks, which belong to deep learning techniques and reflect long-term relationships in time series for a small set of statistical data. The results show that the proposed model can significantly improve the accuracy of forecasts (1–3% of mean absolute percentage error (MAPE) for the analyzed sectors of the economy).

Data Discovery Over Time Series From Star Schemas Based on Association, Correlation, and Causality

This work proposes a methodology applied to repositories modeled using star schemas, such as data marts, to discover relevant time series relations. This paper applies a set of measures related to association, correlation, and causality to create connections among data. In this context, the research proposes a new causality function based on peaks and values that relate coherently time series. To evaluate the approach, the authors use a set of experiments exploring time series about a particular neglected disease that affects several Brazilian cities called American Tegumentary Leishmaniasis and time series about the climate of some cities in Brazil. The authors populate data marts with these data, and the proposed methodology has generated a set of relations linking the notifications of this disease to the variation of temperature and pluviometry.

LSTM Model Based on Multi-Feature Extractor to Detect Flow Pattern Change Characteristics and Parameter Measurement

In the field of petroleum multiphase flow research, accurate measurement of oil-gas two-phase flow parameters is the research focus. This paper analyzes the changing characteristics of the flow pattern of oil-gas two-phase flow and measures the parameters. The experimental phenomenon shows that under different oil and gas flow rate, the flow patterns upstream and downstream of the venturi tube will change to varying degrees. The Convolutional Neural Network often used in the study of flow patterns, but the flow pattern recognition by the CNN model cannot consider the time series relationship. Therefore, this paper proposes a multi-feature extractor-based LSTM (MFE-LSTM) model to predict the parameters. First, the flow pattern features are extracted through the CNN model, and then the features extracted from different CNN models are combined to form a sentence. The relationship between different words in a sentence and different sentences is explored through the LSTM model. The MFE-LSTM algorithm analyzes the changing relationship between flow patterns with time series and spatial position. In addition, in this paper, the flow pattern horizontal splicing model and the flow pattern channel superposition model are designed. In this experiment, the distribution of oil flow rate is 1- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$10~ m^{3}/h $ </tex-math></inline-formula> , the distribution of gas flow rate is 20- <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$150~ m^{3}/h $ </tex-math></inline-formula> , and the distribution of GVF (gas void fraction) is 0.25-0.95. The experimental results show that the MFE-LSTM model has the best prediction effect. The average prediction relative error of gas flow rate is 0.387%. The average prediction relative error of oil flow rate is 2.081%.

Medical Treatment Migration Prediction Based on GCN via Medical Insurance Data.

Nowadays, prediction for medical treatment migration has become one of the interesting issues in the field of health informatics. This is because the medical treatment migration behavior is closely related to the evaluation of regional medical level, the rational use of medical resources, and the distribution of medical insurance. Therefore, a prediction model for medical treatment migration based on medical insurance data is introduced in this paper. First, a medical treatment graph is constructed based on medical insurance data. The medical treatment graph is a heterogeneous graph, which contains entities such as patients, diseases, hospitals, medicines, hospitalization events, and the relations between these entities. However, existing graph neural networks are unable to capture the time-series relationships between event-type entities. To this end, a prediction model based on Graph Convolutional Network (GCN) is proposed in this paper, namely, Event-involved GCN (EGCN). The proposed model aggregates conventional entities based on attention mechanism, and aggregates event-type entities based on a gating mechanism similar to LSTM. In addition, jumping connection is deployed to obtain the final node representation. In order to obtain embedded representations of medicines based on external information (medicine descriptions), an automatic encoder capable of embedding medicine descriptions is deployed in the proposed model. Finally, extensive experiments are conducted on a real medical insurance data set. Experimental results show that our model's predictive ability is better than the best models available.

Volatility spillovers and hedging effectiveness between the oil market and Eurozone sectors: A tale of two crises

This paper aims to analyze the mean and volatility spillovers between oil prices and the Eurozone supersector returns. It uses daily data of the Brent prices and 19 Eurozone supersector indices for the period from August 2004 to August 2015. This area experienced two important instabilities in that period, the global financial crisis (GFC) and the Euro debt crisis (EDC). Because financial turbulences are suspected to induce changes in the volatility dynamics, the full sample is divided into three sub-samples. Empirically, this study employs a bivariate VAR-BEKK-GARCH model that allows for transmission in volatility. The obtained volatilities and covariances are used to compute the optimal weights and hedge ratios for oil–stock portfolio holdings. The findings show that both mean and volatility spillovers between the oil market and the different Eurozone sectors are time-varying and heterogeneous. In the GFC sub-period, there is evidence of contagion effects because there is an intensification of volatility spillovers. The EDC does not seem to have induced any particular change in the spillover effects. The optimal weights, hedge ratios, and correlation analysis results allow an accurate understanding of the time series relationship between the two markets and are useful for financial market participants and policymakers.

Short-term Load Forecasting Model Based on IBFO-BILSTM

Short-term power load forecasting has always been a very important position in power grid operation. Improving prediction accuracy has great significance for improving the power company’s dispatch efficiency and economic benefits. In view of the feed-forward neural network cannot learn the time series relationship between load data, a hybrid depth model based on improved bacterial foraging algorithm (IBFO) and long short-term memory (LSTM) network is proposed for short-term power load forecasting. Using historical load data and its associated temperature data and date information as input, the initial weights and bias are optimized by improved bacterial foraging algorithm, and the multi-layer bidirectional LSTM neural network is used to mine the hidden time series relationship between input data, and finally output prediction. The proposed prediction model was verified by using the historical load data of Nantong. The data shows that the proposed model has higher prediction accuracy than other models.

SOPNet Method for the Fine-Grained Measurement and Prediction of Precipitation Intensity Using Outdoor Surveillance Cameras

Surveillance cameras have been widely used in urban environments and are increasingly used in rural ones. Such cameras have mostly been used for security, but they can be applied to the problem of furnishing fine-grained measurements and predictions of precipitation intensity. In this study, we formulated a stacked order-preserving (OP) learning framework to train a network using time-series data. We constructed an OP module, which uses a three-dimensional (3D) convolution operation to extract features with spatial and temporal information and that are associated with ConvLSTM; this feature extraction is used to learn the short-term and OP time-series relationships between features. Furthermore, the OP modules are stacked to form a stacked OP network (SOPNet), which strengthens the relationship between features in long-term time-series image sequences. This SOPNet can be use to obtain fine-grained measurements and predictions of precipitation intensity from images captured by outdoor surveillance cameras. Our main contributions are threefold. First, the SOPNet strengthens the short-term and long-term time-series relationship between features. Second, the SOPNet simultaneously examines spatial and temporal information to measure and predict precipitation intensity. Third, we constructed a precipitation intensity database based on optical images captured by outdoor surveillance cameras. We experimentally evaluated our proposed architecture using our self-collected data set. We found that SOPNet yields better performance and greater accuracy relative to its well-known state-of-the-art counterparts with respect to various metrics.

Hybrid Deep Learning-Based Model for Wind Speed Forecasting Based on DWPT and Bidirectional LSTM Network

Accurate wind speed forecasting is essential for the reliability and security of the power system, and optimal operation and management of wind integrated smart grids. However, it is still a challenging task due to the highly uncertain and volatile nature of wind speed. Accordingly, in this work, a novel deep learning-based model integrating the discrete wavelet packet transform (DWPT) and bidirectional long short-term memory (BLSTM) is developed to precisely capture deep temporal features and learn the time-varying relationship of wind speed time series. In the proposed method, by applying the DWPT, both approximations and details parts are decomposed by passing through the filters to choose the frequency band related to the features of the original signal more adaptively. The BLSTM networks are incorporated to deal with the uncertainties more effectively as they have bidirectional memory capability (feedforward and feedback loops) to investigate both previous and future hidden layers data. To simultaneously improve the forecasting performance and decrease the learning complexity, the reconstructed state space of historical wind data is employed to reflect the evolution laws of wind speed. Two case studies using real-world wind speed datasets gathered from Flatirons campus (M2) of National Renewable Energy Laboratory (NREL) located in Colorado, USA and weather station of Edmonton, Canada are implemented to demonstrate the effectiveness and superiority of the proposed hybrid method compared to the shallow architectures and state-of-the-art deep learning models in the recent literature.

Driver Lane Change Intention Recognition of Intelligent Vehicle Based on Long Short-Term Memory Network

Driving intention prediction is one of the key technologies for the development of advanced assisted driving systems (ADAS), which could greatly reduce traffic accidents caused by lane change and ensure driving safety. In this paper, an advanced predictive method based on Multi-LSTM (Long Short-Term Memory) is proposed to predict lane change intention effectively. First, the training data set and test set based on real road information data set NGSIM (Next Generation SIMulation) are built considering ego vehicle driving state and the influence of surrounding vehicles. Second, the Multi-LSTM-based prediction controller is constructed to learn vehicle behavior characteristics and time series relation of various states in the process of lane change. Then, the influences of prediction model structure change and data structure change on test results are verified. Finally, the verification tests based on HIL (Hardware-in-the-Loop) simulation are constructed. The results show that the proposed prediction model can accurately predict the vehicle lane change intention in highway scenarios and the maximum prediction accuracy can reach 83.75%, which is higher than that of common method SVM (Support Vector Machine).

Transforming Finance Into Vision: Concurrent Financial Time Series as Convolutional Nets.

We present a novel representation for multiple synchronized financial time series as images, motivated by deep learning methods in machine vision. The research pursues two related strands of inquiry. The first is to transform concurrent synchronized time series analysis-one that is prevalent in Finance and other domains-into a machine vision problem so that the standard deep learning machinery such as convolutional nets can be applied to the transformed problem. The second line of inquiry pursues the idea of transfer learning, where learning occurs on synthetic simulated data corresponding to a finite set of lead-lag relationships in the concurrent time series, and the learned model is applied out of the box to the application domain, in our case, Finance. The successful application of transfer learning, however, requires that a relationship exists between the simulated and real-world data that the learner is able to discern. This relationship helps to bias the learner toward learning things that will be useful in the application domain. We demonstrate the application of our trained model for identifying data-driven regime shifts in financial time series data. We present an analysis of the results and discuss some of the useful properties of the representation and directions for future research.

Deep Learning Based RF Fingerprint Identification Using Differential Constellation Trace Figure

This paper proposes a novel deep learning-based radio frequency fingerprint (RFF) identification method for internet of things (IoT) terminal authentications. Differential constellation trace figure (DCTF), a two-dimensional (2D) representation of differential relationship of signal time series, is utilized to extract RFF features without requiring any synchronization. A convolutional neural network (CNN) is then designed to identify different devices using DCTF features. Compared to the existing CNN-based RFF identification methods, the proposed DCTF-CNN possesses the merits of high identification accuracy, zero prior information and low complexity. Experimental results have demonstrated that the proposed DCTF-CNN can achieve an identification accuracy as high as 99.1% and 93.8% under SNR levels of 30 dB and 15 dB, respectively, when classifying 54 target ZigBee devices, which significantly outperforms the existing RFF identification methods.

Prediction in Autism by Deep Learning Short-Time Spontaneous Hemodynamic Fluctuations.

This study aims to explore the possibility of using a multilayer artificial neural network for the classification between children with autism spectrum disorder (ASD) and typically developing (TD) children based on short-time spontaneous hemodynamic fluctuations. Spontaneous hemodynamic fluctuations were collected by a functional near-infrared spectroscopy setup from bilateral inferior frontal gyrus and temporal cortex in 25 children with ASD and 22 TD children. To perform feature extraction and classification, a multilayer neural network called CGRNN was used which combined a convolution neural network (CNN) and a gate recurrent unit (GRU), since CGRNN has a strong ability in finding characteristic features and acquiring intrinsic relationship in time series. For the training and predicting, short-time (7 s) time-series raw functional near-infrared spectroscopy (fNIRS) signals were used as the input of the network. To avoid the over-fitting problem and effectively extract useful differentiation features from a sample with a very limited size (e.g., 25 ASDs and 22 TDs), a sliding window approach was utilized in which the initially recorded long-time (e.g., 480 s) time-series was divided into many partially overlapped short-time (7 s) sequences. By using this combined deep-learning network, a high accurate classification between ASD and TD could be achieved even with a single optical channel, e.g., 92.2% accuracy, 85.0% sensitivity, and 99.4% specificity. This result implies that the multilayer neural network CGRNN can identify characteristic features associated with ASD even in a short-time spontaneous hemodynamic fluctuation from a single optical channel, and second, the CGRNN can provide highly accurate prediction in ASD.

Time Series Analysis for the Stock Repurchases and Issuance

본 연구는 1994-2017년 유가증권시장에 보고된 자사주매입과 증자의 자료를 이용하여 두 자본거래의 시계열 패턴에 대한 공통적 설명 요인을 검증하였다. 설명 요인으로 제시된 것은 마켓타이밍과 경기순환주기이다. 검증은 변수 간의 공적분을 감안하기 위한 2단계 오차수정모형을 이용하였다. 검증 결과, 마켓타이밍 변수는 일부 자사주매입 모형을 제외하면 유의한 설명력을 가지지 못하는 것으로 나타났다. 반면 경기순환주기는 자사주매입과 증자 모두에 대해 유의한 설명력을 가지는 것으로 검증되었다. 경기순환이라는 거시적 요인이 기업의 현금흐름과 투자에 영향을 미쳐 자본거래를 인과할 것이라는 가설과 일관된 결과이다. 자사주매입과 증자의 유행시기가 유사한 것도 경기순환주기 가설에는 부합하나 마켓타이밍 이론에는 어긋나는 증거이다. 다만, 현금흐름과 투자지출 변수의 검증 결과는 경기순환주기 가설에 완전히 부합하지는 않았다. 본 연구는 횡단면 분석에 치중되었던 기존 연구들과 달리 자본거래의 시계열을 분석 범위로 설정했다는 점에서 의미를 찾을 수 있으며, 경기순환주기라는 거시 변수를 두 자본거래의 공통적 설명요인으로 제시했다는 점에서 마켓타이밍에 의존해 왔던 기존 연구들과 차별성을 지닌다.This study examines the market timing and business cycle hypotheses for the time series patterns of stock repurchases and SEOs (seasoned equity offerings). The sample consists of stock repurchases and SEOs reported by the firms listed on the KOSPI during 1994-2017. The time series relation between variables is examined using the two-step error correction model. The result shows that waves of the two capital transactions were in line, which is unexplainable by market-timing theory but consistent with business cycle hypothesis. Market timing variables do not have significant explanatory power. On the contrary, the business cycle variable has significant explanatory power on each time-series pattern of stock repurchases and SEOs. The result suggests that the business cycle is a common causal factor for the two capital transactions. This study helps to understand the corporate capital transactions by analyzing their time-series patterns unlike extant studies that have focused on cross-sectional analysis.

Comparison of high versus low frequency cerebral physiology for cerebrovascular reactivity assessment in traumatic brain injury: a multi-center pilot study

Current accepted cerebrovascular reactivity indices suffer from the need of high frequency data capture and export for post-acquisition processing. The role for minute-by-minute data in cerebrovascular reactivity monitoring remains uncertain. The goal was to explore the statistical time-series relationships between intra-cranial pressure (ICP), mean arterial pressure (MAP) and pressure reactivity index (PRx) using both 10-s and minute data update frequency in TBI. Prospective data from 31 patients from 3 centers with moderate/severe TBI and high-frequency archived physiology were reviewed. Both 10-s by 10-s and minute-by-minute mean values were derived for ICP and MAP for each patient. Similarly, PRx was derived using 30 consecutive 10-s data points, updated every minute. While long-PRx (L-PRx) was derived via similar methodology using minute-by-minute data, with L-PRx derived using various window lengths (5, 10, 20, 30, 40, and 60 min; denoted L-PRx_5, etc.). Time-series autoregressive integrative moving average (ARIMA) and vector autoregressive integrative moving average (VARIMA) models were created to analyze the relationship of these parameters over time. ARIMA modelling, Granger causality testing and VARIMA impulse response function (IRF) plotting demonstrated that similar information is carried in minute mean ICP and MAP data, compared to 10-s mean slow-wave ICP and MAP data. Shorter window L-PRx variants, such as L-PRx_5, appear to have a similar ARIMA structure, have a linear association with PRx and display moderate-to-strong correlations (r ~ 0.700, p < 0.0001 for each patient). Thus, these particular L-PRx variants appear closest in nature to standard PRx. ICP and MAP derived via 10-s or minute based averaging display similar statistical time-series structure and co-variance patterns. PRx and L-PRx based on shorter windows also behave similarly over time. These results imply certain L-PRx variants may carry similar information to PRx in TBI.

Macroeconomic uncertainty, information competition, and liquidity

In this study we provide evidence regarding the effect of macroeconomic uncertainty on stock market liquidity by analyzing the time-series relation between both types of macroeconomic uncertainty shocks: good and bad uncertainty and various measures of liquidity. Our study also sheds light on whether the impact of macroeconomic uncertainty on liquidity varies with information environment by examining the effects of investor competition over information on the relation between good and bad uncertainty and liquidity in the Taiwan markets. We find that there is a negative (positive) relation between bad (good) uncertainty and the subsequent liquidity of stocks. We also find that the magnitude of the impact of bad uncertainty on stock market liquidity is larger than good uncertainty. On the other side, macroeconomic uncertainty would exerts a larger impact on liquidity when the market is imperfectly competitive.

A Novel Aspect-based Sentiment Analysis Network Model Based on Multilingual Hierarchy in Online Social Network

AbstractIn recent years, sentiment analysis based on aspects has become one of the research hotspots in the field of natural language processing. Aiming at the fact that the existing network model cannot fully obtain the interrelationship between sentences in the same comment and the long-distance dependence of specific aspects in the whole comment, a multilingual deep hierarchical model combining regional convolutional neural network and bidirectional LSTM network is proposed. The model obtains the time series relationship of different sentences in the comments through the regional CNN, and obtains the local features of the specific aspects in the sentence and the long-distance dependence in the whole comment through the hierarchical attention network. In addition, the model improves the word vector representation based on the gate mechanism to make the model completely independent of the language. Experimental results for different domain datasets in multi-language show that the proposed model achieves better classification results than the traditional deep network model, the network model combining with the attention mechanism and considering the relationship between sentences.

Detection of Lead-Lag Relationships Using Both Time Domain and Time-Frequency Domain; An Application to Wealth-To-Income Ratio

The objective of this paper is the joint application of two different methodological concepts for the detection of lead-lag relationships in economic time-series in order to investigate their consistency and their potential complementarity. The first methodology, a time domain analysis based on vector error correction model, provides evidence about the existence of long-run equilibrium of the time-series and the short-run lead-lag behaviors. The second methodology, a time-frequency concept based on the phase difference of the cross-wavelet coherence, analyzes the lead-lag relationships across various timescales and reveals the altering of leadership over time. The two methods are applied to time-series of wealth-to-income ratio of four developed countries over the period 1970–2010 and analyze the lead-lag relationships of the countries in the long-run and in the short-run. The results show that the two methods are consistent in their major long-run findings, however, they reveal different aspects regarding the short-run dynamics of the lead-lag relationships. Furthermore, the results suggest the complementarity of the two methodologies in the context of a complete framework for the analysis of the lead-lag relationships in non-stationary economic time-series.

A correlation based feature representation for first-person activity recognition

In this paper, a simple yet efficient activity recognition method for first-person video is introduced. The proposed method is appropriate for representation of high-dimensional features such as those extracted from convolutional neural networks (CNNs). The per-frame (per-segment) extracted features are considered as a set of time series, and inter and intra-time series relations are employed to represent the video descriptors. To find the inter-time relations, the series are grouped and the linear correlation between each pair of groups is calculated. The relations between them can represent the scene dynamics and local motions. The introduced grouping strategy helps to considerably reduce the computational cost. Furthermore, we split the series in temporal direction in order to preserve long term motions and better focus on each local time window. In order to extract the cyclic motion patterns, which can be considered as primary components of various activities, intra-time series correlations are exploited. The representation method results in highly discriminative features which can be linearly classified. The experiments confirm that our method outperforms the state-of-the-art methods on recognizing first-person activities on the two challenging first-person datasets.

Returns to Scale Among Corporate Bond Mutual Funds

I document a (within-fund) hump-shaped relation between fund size and subsequent fund performance among U.S. corporate bond mutual funds. When funds are small, they exhibit increasing returns to scale but when they become large, they exhibit decreasing returns to scale. This sharply contrasts with the previous finding of decreasing returns to scale among equity mutual funds. Further, I show that the nature of trading cost in the corporate bond market --- in particular, a U-shaped relation between trade size and unit trading cost at the corporate bond level --- is relevant for explaining hump-shaped returns to scale. Interpreting these empirical patterns is not straightforward, though. In a rational expectations framework, we expect a fund's net alpha always to be zero and hence, no time-series relation between fund size and subsequent fund alpha. To help interpret the empirical findings, I propose a dynamic model in which investors learn about a fund's ability to manage its trading cost from its past returns. The evolution of investors' beliefs provides a source of variation in fund size and further, in fund alpha in equilibrium over time.