Multiple Extreme Learning Machines Based Arrival Time Prediction for Public Bus Transport

Due to fast-growing urbanization,the traffic management system becomes a crucial problem owing to the rapid growth in the number of vehicles The research proposes an Intelligent public transportation system where informa-tion regarding all the buses connecting in a city will be gathered,processed and accurate bus arrival time prediction will be presented to the user.Various linear and time-varying parameters such as distance,waiting time at stops,red signal duration at a traffic signal,traffic density,turning density,rush hours,weather conditions,number of passengers on the bus,type of day,road type,average vehi-cle speed limit,current vehicle speed affecting traffic are used for the analysis.The proposed model exploits the feasibility and applicability of ELM in the travel time forecasting area.Multiple ELMs(MELM)for explicitly training dynamic,road and trajectory information are used in the proposed approach.A large-scale dataset(historical data)obtained from Kerala State Road Transport Corporation is used for training.Simulations are carried out by using MATLAB R2021a.The experiments revealed that the efficiency of MELM is independent of the time of day and day of the week.It can manage huge volumes of data with less human intervention at greater learning speeds.It is found MELM yields prediction with accuracy in the range of 96.7%to 99.08%.The MAE value is between 0.28 to 1.74 minutes with the proposed approach.The study revealed that there could be regularity in bus usage and daily bus rides are predictable with a better degree of accuracy.The research has proved that MELM is superior for arrival time pre-dictions in terms of accuracy and error,compared with other approaches.

Ship motion extreme short time prediction of ship pitch based on diagonal recurrent neural network

A DRNN (diagonal recurrent neural network) and its RPE (recurrent prediction error) learning algorithm are proposed in this paper .Using of the simple structure of DRNN can reduce the capacity of calculation. The principle of RPE learning algorithm is to adjust weights along the direction of Gauss-Newton. Meanwhile, it is unnecessary to calculate the second local derivative and the inverse matrixes, whose unbiasedness is proved. With application to the extremely short time prediction of large ship pitch, satisfactory results are obtained. Prediction effect of this algorithm is compared with that of auto-regression and periodical diagram method, and comparison results show that the proposed algorithm is feasible.

Estimation of Dynamic VaR in Chinese Stock Markets Based on Time Scale and Extreme Value Theory

The accuracy and time scale invariance of value-at-risk （VaR） measurement methods for different stock indices and at different confidence levels are tested. Extreme value theory （EVT） is applied to model the extreme tail of standardized residual series of daily/weekly indices losses, and parametric and nonparametric methods are used to estimate parameters of the general Pareto distribution （GPD）, and dynamic VaR for indices of three stock markets in China. The accuracy and time scale invariance of risk measurement methods through back-testing approach are also examined. Results show that not all the indices accept time scale invariance; there are some differences in accuracy between different indices at various confidence levels. The most powerful dynamic VaR estimation methods are EVT-GJR-Hill at 97.5% level for weekly loss to Shanghai stock market, and EVT-GARCH-MLE （Hill） at 99.0% level for weekly loss to Taiwan and Hong Kong stock markets, respectively.

Real Time Monitoring of Extreme Rainfall Events with Simple X-Band Mini Weather Radar

Real time rainfall events monitoring is very important for a large number of reasons: Civil Protection, hydrogeological risk management, hydroelectric power purposes, road and traffic regulation, and tourism. Efficient monitoring operations need continuous, high-resolution and large-coverage data. To monitor and observe extreme rainfall events, often much localized over small basins of interest, and that could frequently causing flash floods, an unrealistic extremely dense rain gauge network should be needed. On the other hand, common large C-band or S-band long range radars do not provide the necessary spatial and temporal resolution. Simple short-range X-band mini weather radar can be a valid compromise solution. The present work shows how a single polarization, non-Doppler and non-coherent, simple and low cost X-band radar allowed monitoring three very intense rainfall events occurred near Turin during July 2014. The events, which caused damages and floods, are detected and monitored in real time with a sample rate of 1 minute and a radial spatial resolution of 60 m, thus allowing to describe the intensity of the precipitation on each small portion of territory. This information could be very useful if used by authorities in charge of Civil Protection in order to avoid inconvenience to people and to monitor dangerous situations.

Features of aerosol optical depth and its relation to extreme temperatures in China during 1980–2001

Based on Total Ozone Mapping Spectrometer （TOMS） monthly aerosol optical thickness （AOT） measurements in 1980–2001 a study is made of space/time patterns and difference between land and sea of AOT 0.50 μm thick over China,which are put into correlation analysis with synchronous extreme temperature indices （warm/cold day and night）.Results suggest that 1） the long-term mean AOT over China is characterized by typical geography,with pronounced land-sea contrast.And AOT has significant seasonality and its seasonal difference is diminished as a function of latitude.2） On the whole,the AOT displays an appreciably increasing trend,with the distinct increase in the eastern Qinghai-Tibetan plateau and SW China,North China,the mid-lower Changjiang （MiLY） valley as well as the South China Sea,but marginal decrease over western/northern Xinjiang and part of South China.3） The AOT over land and sea is marked by conspicuous intra-seasonal and -yearly oscillations,with remarkable periods at one-,two-yr and more （as interannual periods）.4） Land AOT change is well correlated with extremely temperature indexes.Generally,the correlations of AOT to the extreme temperature indices are more significant in Eastern China with 110 ° E as the division.Their high-correlation regions are along the Southern China coastline,the Loess Plateau and the Sichuan Basin,and even higher in North China Plain and the mid-lower Changjiang River reaches.5） Simulations of LMDZ-regional model indicate that aerosol effects may result in cooling all over China,particularly in Eastern China.The contribution of aerosol change may result in more decrease in the maximum temperature than the minimum,with decrease of 0.11/0.08 K for zonal average,respectively.

Statistics of extreme events in Chinese stock markets

We investigate the impact of financial factors on daily volume recurrent time intervals in the developing Chinese stock markets. The tails of probability distribution functions（PDFs） of volume recurrent intervals behave as a power-law, and the scaling exponent decreases with the increase of stock lifetime, which are similar to those in the US stock markets, and they are typical representatives of developed markets. The difference is that the power-law exponent values remain almost the same with the changes of market capitalization, mean volume, and mean trading value, respectively. These findings enrich the results for event statistics for financial markets.

Extreme learning with chemical reaction optimization for stock volatility prediction

Extreme learning machine(ELM)allows for fast learning and better generalization performance than conventional gradient-based learning.However,the possible inclusion of non-optimal weight and bias due to random selection and the need for more hidden neurons adversely influence network usability.Further,choosing the optimal number of hidden nodes for a network usually requires intensive human intervention,which may lead to an ill-conditioned situation.In this context,chemical reaction optimization(CRO)is a meta-heuristic paradigm with increased success in a large number of application areas.It is characterized by faster convergence capability and requires fewer tunable parameters.This study develops a learning framework combining the advantages of ELM and CRO,called extreme learning with chemical reaction optimization(ELCRO).ELCRO simultaneously optimizes the weight and bias vector and number of hidden neurons of a single layer feed-forward neural network without compromising prediction accuracy.We evaluate its performance by predicting the daily volatility and closing prices of BSE indices.Additionally,its performance is compared with three other similarly developed models—ELM based on particle swarm optimization,genetic algorithm,and gradient descent—and find the performance of the proposed algorithm superior.Wilcoxon signed-rank and Diebold–Mariano tests are then conducted to verify the statistical significance of the proposed model.Hence,this model can be used as a promising tool for financial forecasting.

A FAST ALGORITHM FOR EXTREME FILTER OF 2D IMAGE

ＡＦＡＳＴＡＬＧＯＲＩＴＨＭＦＯＲＥＸＴＲＥＭＥＦＩＬＴＥＲＯＦ２ＤＩＭＡＧＥＸｕＬｅｉ（许磊）ＬｉＪｉｅｇｕ（李介谷）ＬｉＷｅｎｓｈｕｎ（李文舜）（ＩｎｓｔｉｔｕｔｅｏｆＰａｔｅｒｎＲｅｃｏｇｎｉｔｉｏｎ＆ＩｍａｇｅＰｒｏｃｅｓｓｉｎｇ，Ｓｈａｎｇ...

EXTREMA OF A GAUSSIAN RANDOM FIELD:BERMAN'S SOJOURN TIME METHOD

In this paper we devote ourselves to extending Berman’s sojourn time method,which is thoroughly described in[1-3],to investigate the tail asymptotics of the extrema of a Gaussian random field over[0,T]^(d) with T∈(0,∞).

Applications of Bootstrap in Analyzing General Extreme Value Distributions

The bootstrap method is one of the new ways of studying statistical math which this article uses but is a major tool for studying and evaluating the values of parameters in probability distribution.Our research is concerned overview of the theory of infinite distribution functions.The tool to deal with the problems raised in the paper is the mathematical methods of random analysis(theory of random process and multivariate statistics).In this article,we introduce the new function to find out the bias and standard error with jackknife method for Generalized Extreme Value distributions.

Raman scattering under extreme conditions

Raman scattering is a versatile and powerful technique and has been widely used in modern scientific research and vast industrial applications. It is one of the fundamental experimental techniques in condensed matter physics, since it can sensitively probe the basic elementary excitations in solids like electron, phonon, magnon, etc. The application of extreme conditions （low temperature, high magnetic field, high pressure, etc.） to Raman scattering, will push its capability up to an unprecedented level, because this enables us to look into new quantum phases driven by extreme conditions, trace the evolution of the excitations and their coupling, and hence uncover the underlying physics. This review contains two topics. In the first part, we will introduce the Raman facility under extreme conditions, belonging to the optical spectroscopy station of Synergetic Extreme Condition User Facilities （SECUF）, with emphasis on the system design and the capability the facility can provide. Then in the second part we will focus on the applications of Raman scattering under extreme conditions to a variety of condensed matter systems such as superconductors, correlated electron systems, charge density waves （CDW） materials, etc. Finally, as a rapidly developing technique, time-resolved Raman scattering will be highlighted here.

基于多因素分析的烘丝机智能调控研究

烘丝过程中影响出口烟丝水分和温度的干扰因素较多,为了保障出口烟丝水分与温度的稳定性,首先基于烘丝过程的历史数据,采用相关性分析方法对影响因子进行筛选。其次,选取出口烟丝水分和出口烟丝温度为目标值,通过对各机器学习模型的对比分析选取了能够快速建模与预测的极限学习机(extreme learning machine,ELM)作为建模方法,以通过模型求解运算,得出预测值。最后,采用模拟退火(simulated annealing,SA)算法,实时优化热风风速和排潮风门开度的设定值,实现对出口烟丝水分和温度的预测和控制。实验结果表明,极限学习机模型的预测效果良好,预测当前出口水分的均方根误差为0.015,出口温度的均方根误差为0.638,误差较小,保障了烘丝机智能调控方法的调控精度。

iFlow彩色血流编码成像技术在下肢动脉硬化闭塞症诊断中的应用价值

目的研究iFlow彩色血流编码成像技术在下肢动脉硬化闭塞症(LEASO)诊断中的应用价值。方法选择2022年3月至2023年10月期间确诊的106例LEASO患者作为本研究的LEASO组,以一般资料与LEASO组匹配且无动脉病变的80例志愿者作为对照组。两组受试者均进行数字减影血管造影(DSA)并采用iFlow彩色血流编码成像技术检测股骨头区域和踝关节区域达峰时间(TTP)、计算踝关节区域与股骨头区域TTP的差值,测量踝肱指数(ABI)。结果两组研究对象年龄、性别、体质量指数、吸烟史、高血压病史、糖尿病史、冠心病史、股骨头区域TTP的比较,差异无统计学意义(P>0.05);LEASO组踝关节区域TTP及TTP差值均高于对照组,差异有统计学意义(P<0.05);LEASO组中不同Rutherford分类患者股骨头区域TTP的比较以及左侧病变患者与右侧病变患者股骨头区域TTP、踝关节区域TTP、TTP差值的比较,差异无统计学意义(P>0.05),Rutherford分类越高,踝关节区域TTP及TTP差值越低(P<0.05);经Pearson检验,LEASO患者的踝关节区域TTP、TTP差值与ABI呈负相关(P<0.05);经受试者工作特征(ROC)曲线分析,踝关节区域TTP、TTP差值对LEASO具有诊断效能;经Delong检验,TTP差值诊断的ROC曲线下面积高于踝关节区域TTP(P<0.05)。结论iFlow彩色血流编码成像技术测定踝关节区TTP及TTP差值是诊断LEASO的量化指标。

太行山复杂地形下华北暖季极端降水的时空分布特征

基于2012—2021年5—9月华北五省的逐日降水资料和台站地形高度数据,统计分析了华北全区及各子区域极端降水事件的降水量及其强度和频次的时空分布特征;并运用地理加权回归(GWR)模型分析得到极端降水事件的降水量、强度及频次与海拔高度之间的关系。结果表明:1)华北区域极端降水量的时间变化均呈多波动特征且区域差异性显著,太行山以西高原和以东平原降水频次多、波动明显且强度较弱,太行山南段以南平原降水频次少、变化平缓而强度明显偏强。2)极端降水量的空间分布呈现南北少、中间多的型态分布,降水量大值区分别位于燕山东南侧和太行山南段晋冀豫三省交界处;极端降水高频站点主要聚集在晋东南地区;日最大降水量超过300 mm的站点主要集中在太行山脉和燕山山脉与华北平原的过渡地带。3)华北区域38°N以北,极端降水量、降水频次、强度和日最大降水量均随海拔高度的升高而减小;38°N以南,山西南部临运地区降水量随海拔高度的升高而显著增加。由于降水频次和强度与地形均存在正相关而导致,太行山附近降水量随海拔高度的升高而减小的贡献主要在于降水强度而非降水频次。

四川两次极端暴雨强降水特征及与雷达回波和闪电关系分析

选取了2020年8月四川两次历史性极端暴雨过程,根据量级和持续时间对强降水进行划分,分析其时空分布特征及与雷达回波和闪电的关系。结果表明:短时强降水主要发生在22时—次日03时,降水强度为30~<50 mm/h的站次最多,区域集中在盆地西部。随着降水量级的增加,站点对应的闪电密度均增大,小时平均回波、最强回波、最弱回波均呈增强的趋势。随着降水持续时间的增加,站点对应的负地闪平均强度增强。第一次过程强降水站次与闪电频次的高值中心具有良好的对应关系。第二次过程随着降水量级增大,对应的回波均方根误差减小,而第一次过程则相反。

基于极限梯度提升和探地雷达时频特征的水泥路面脱空识别

针对探地雷达(GPR)数据解译依赖于人工经验,存在费时费力和主观偏差的问题,提出了基于极限梯度提升(XGBoost)和GPR时频特征的水泥路面脱空识别方法。采用正演模拟、室内试验和现场试验获得了脱空病害数据源,建立含有标签的脱空GPR数据集;通过重采样方法统一GPR数据采样频率,并对预处理后的GPR数据进行时频域特征提取,建立了包含18个时域和12个频域特征的数据集。以时频域特征为输入,是否存在脱空病害为输出,采用XGBoost算法构建脱空识别模型,并与随机森林(RF)和人工神经网络(ANN)算法进行对比。结果表明,模型的识别准确率排序为XGBoost(98.10%)>ANN(95.10%)>RF(93.17%),XGBoost模型识别精度最高,并能在实际路面上准确定位脱空区域。

基于改进极限学习机的电力市场实时电价预测方法

针对实时电价数据波动性强及其影响因素复杂,导致现有预测模型稳定性及预测精度低的问题,提出了一种基于改进极限学习机的电力市场实时电价预测方法。设计了基于经验小波变换的实时电价数据分解方法,将电价序列分解为接近价格的低频信号和噪声的高频信号。同时提出基于改进随机森林的实时电价特征提取算法,根据预测重要度获取最优的电价影响因素特征组合。以此为基础,将核函数替代极限学习机隐藏层构建了R-KELM预测模型,更好地反映了多因素影响下实时电价的不确定性和波动性。以PJM实时电价数据为例,结果表明,所提方法可以有效克服电价数据强波动性及高特征冗余的问题,预测模型准确性及稳定性得到显著提升。

沿海地区良态风和台风气候不同时距风速转换系数研究

以华南地区某海岸机场区域2012—2021年的风速实测数据为对象,根据这10 a内台风的登陆情况,将其划分为良态风气候和台风气候,研究不同风气候环境下风速时距转换系数的变化规律。研究结果表明:在良态风气候下,月、季、年极值风速的分布规律均符合极值Ⅰ型(Gumbel分布),统计发现转换系数随时距增大呈指数递减的变化规律,基于此拟合得到良态风气候下月、季、年极值风速的时距转换系数公式;在台风气候下,时距转换系数曲线的平稳程度随台风强度的减小而减弱;时距3 s与10 min的转换系数的分布规律符合极值Ⅰ型分布,30 min、1 h与10 min的转换系数的分布规律符合极值Ⅲ型分布;使用良态风转换系数公式估计台风气候时距3 s与10 min的转换系数时,会存在10%~30%的相对误差,可将转换系数乘以对应的放大系数去近似估计,或直接采用1.63的转换系数。该研究结果可为不同风气候条件的风荷载计算提供参考。

高比例新能源接入下的历史策略库辅助源网荷储协同实时电压控制研究

随着“双碳”目标的推进,储能、充电桩等用户侧新型负荷设备数量增多。在此背景下,由于量测条件不满足造成的不可观测区域将直接导致传统的电压控制方法难以完成对分布式电源的精准调控。为解决上述问题,提出一种基于蜣螂优化算法(DBO)的极限学习机(ELM)构建历史策略库,用以辅助源网荷储协同实时电压控制的方法,可实现对配电网电压实时精确控制。首先介绍了基于近似灵敏度计算的电压控制方法,然后介绍了DBO改进的极限学习机和历史策略库的概念及结合应用方法,构建了以基于近似灵敏度计算的电网节点有功及无功功率为输入,母线期望电压为输出的ELM模型。模型输出的母线电压作为控制依据,进一步转换为下发的用户侧可调设备调节指令。仿真算例的结果验证了所提方法的有效性和优越性。