基于CEEMDAN-VMD-TCN-lightGBM模型的水质预测研究

针对目前水质预测模型中因为数据本身的复杂性、在信号处理过程中存在的噪声干扰以及分解深度不够导致单一分解难以全面捕捉信号非线性特征的问题,提出了一种基于二次分解的水质预测模型。该模型采用完全自适应噪声集合经验模态分解(CEEMDAN)对原始数据进行分解,再利用变分模态分解(VMD)对熵值最高的模态分量进行二次分解,最终将处理后的时间序列输入到TCN-lightGBM多特征预测模型中。同时,采用麻雀算法(SSA)对预测模型进行优化。以山东省玉符河水质为例,本模型的均方根误差(RMSE)是0.1053,平均绝对误差(MAE)是0.0815,决定系数(R2)是0.9471,与GRU、LSTM、LightGBM、TCN等当下较为流行的模型的预测指标进行比较。结果显示,在R2上本模型提升了53.04%、70.41%、66.07%、65.20%等,在RMSE上减少了62.76%、65.50%、64.93%、64.80%等,在MAE上降低了62.76%、66.24%、63.80%、65.24%等。由此可知,基于CEEMDAN-VMD-TCN-lightGBM的模型具有更好的预测性能、泛化能力和捕捉信号非线性特征的能力。

沉默RPARP-AS1对GBM U251细胞增殖、迁移和凋亡的影响

目的探究沉默长链非编码RNA RPARP-AS1对胶质母细胞瘤(GBM)U251细胞增殖、迁移和凋亡的影响。方法选取80例GBM组织与癌旁组织。将GBM细胞株U251细胞随机分为Control组、小干扰RNA阴性对照(si-NC)组、si-RPARP-AS1组、si-RPARP-AS1+抑制剂(inhibitor)NC组、si-RPARP-AS1+miR-339-5p inhibitor组。qRT-PCR检测组织和细胞RPARP-AS1、miR-339-5p和鼠双微体基因2(MDM2)mRNA的表达。CCK-8检测各组细胞增殖能力。Transwell检测各组U251细胞迁移和侵袭能力。流式细胞术检测细胞凋亡率。Western blotting检测B细胞淋巴瘤-2(Bcl-2)、Bcl-2相关X蛋白(Bax)、E-钙黏蛋白(E-cadherin)、N-钙黏蛋白(N-cadherin)和MDM2蛋白的表达。双荧光素酶报告基因实验分别验证RPARP-AS1和miR-339-5p、miR-339-5p和MDM2的关系。结果与癌旁组织相比,GBM组织RPARP-AS1、MDM2 mRNA高表达,miR-339-5p低表达(P<0.05)。与Control组、si-NC组相比,si-RPARP-AS1组U251细胞RPARP-AS1、MDM2 mRNA和Bcl-2、N-cadherin、MDM2蛋白表达以及细胞增殖、迁移、侵袭能力显著降低(P<0.05),miR-339-5p、Bax、E-cadherin蛋白表达以及细胞凋亡率显著升高(P<0.05)。与si-RPARP-AS1组、si-RPARP-AS1+inhibitor NC组比较,si-RPARP-AS1+miR-339-5p inhibitor组细胞MDM2 mRNA和Bcl-2、N-cadherin、MDM2蛋白表达以及细胞增殖、迁移、侵袭能力显著升高(P<0.05),miR-339-5p、Bax、E-cadherin蛋白表达以及细胞凋亡率显著降低(P<0.05)。RPARP-AS1靶向负调控miR-339-5p表达,miR-339-5p靶向负调控MDM2表达。结论沉默RPARP-AS1对GBM U251细胞增殖、迁移和侵袭具有抑制作用,对细胞凋亡具有促进作用,可能与通过上调miR-339-5p,抑制MDM2表达有关。

一种基于随机森林和Light GBM的房产估价模型

针对商品房评估方法中存在数据源单一,考虑影响因素理想化,特征工程过分依赖主观经验等问题,结合随机森林和Light GBM模型,提出了一种RF_LightGBM模型用于房产价值评估。首先,通过随机森林对特征进行重要度排序,将影响房产价格因素较小的特征排除,使用网格搜索算法对模型进行优化,最后将该方法用于房产价值评估。在真实的房价数据集上进行的实验表明,相较于随即森林,XGBoost等传统模型,RF_LightGBM模型的评估精度提高了1.7%,且百分误差在0%~10%以内的评估结果占比88.38%。说明所用模型可以很好地应用于房产价值评估,得到的评估结果更加准确。

Pulses in ground motions identified through surface partial matching and their impact on seismic rocking consequence

In seismology and earthquake engineering,it is fundamental to identify and characterize the pulse-like features in pulse-type ground motions.To capture the pulses that dominate structural responses,this study establishes congruence and shift relationships between response spectrum surfaces.A similarity search between spectrum surfaces,supplemented with a similarity search in time series,has been applied to characterize the pulse-like features in pulse-type ground motions.The identified pulses are tested in predicting the rocking consequences of slender rectangular blocks under the original ground motions.Generally,the prediction is promising for the majority of the ground motions where the dominant pulse is correctly identified.

Monitoring absolute vertical land motions and absolute sea-level changes from GPS and tide gauges data over French Polynesia

In this study,we estimate the absolute vertical land motions at three tidal stations with collocated Global Navigation Satellite System(GNSS)receivers over French Polynesia during the period 2007-2020,and obtain,as ancillary results,estimates of the absolute changes in sea level at the same locations.To verify our processing approach to determining vertical motion,we first modeled vertical motion at the International GNSS Service(IGS)THTI station located in the capital island of Tahiti and compared our estimate with previous independent determinations,with a good agreement.We obtained the following estimates for the vertical land motions at the tide gauges:Tubuai island,Austral Archipelago-0.92±0.17 mm/yr,Vairao village,Tahiti Iti:-0.49±0.39 mm/yr,Rikitea,Gambier Archipelago-0.43±0.17 mm/yr.The absolute variations of the sea level are:Tubuai island,Austral Archipelago 5.25±0.60 mm/yr,Vairao village,Tahiti Iti:3.62±0.52 mm/yr,Rikitea,Gambier Archipelago 1.52±0.23 mm/yr.We discuss these absolute values in light of the values obtained from altimetric measurements and other means in French Polynesia.

Maxsurf Motions模块在船舶耐波性上的应用与验证

船舶作为海洋中最主要的运动物体,其在海浪中的运动问题,是船舶适航性、耐波性的基础,同时也是船舶设计建造中的关键步骤。采用Maxsurf Motions模块对4种国际标准船模进行了耐波性计算,验证了Maxsurf Motions中的切片理论和面元法的计算精度与可靠性,探讨了Motions程序在船舶耐波性分析中的可行性,分析了Motions模块计算误差的原因。结果表明:Motions模块对船舶运动响应以及波浪增阻的计算精度基本令人满意。对于附加质量系数和阻尼系数,Motions在中高频波段的计算精度较高。研究成果对Maxsurf Motions的应用起到一定的借鉴和指导作用,同时也提供了一系列耐波性计算的验证算例。

Effect of the mineral spatial distribution heterogeneity on the tensile strength of granite:Insights from PFC3D-GBM numerical analysis

The mechanical characteristics of crystalline rocks are affected by the heterogeneity of the spatial distribution of minerals.In this paper,a novel three-dimensional(3D)grain-based model(GBM)based on particle flow code(PFC),i.e.PFC3D-GBM,is proposed.This model can accomplish the grouping of mineral grains at the 3D scale and then filling them.Then,the effect of the position distribution,geometric size,and volume composite of mineral grains on the cracking behaviour and macroscopic properties of granite are examined by conducting Brazilian splitting tests.The numerical results show that when an external load is applied to a sample,force chains will form around each contact,and the orientation distribution of the force chains is uniform,which is independent of the external load level.Furthermore,the number of high-strength force chains is proportional to the external load level,and the main orientation distribution is consistent with the external loading direction.The main orientation of the cracks is vertical to that of the high-strength force chains.The geometric size of the mineral grains controls the mechanical behaviours.As the average grain size increases,the number of transgranular contacts with higher bonding strength in the region connecting both loading points increases.The number of high-strength force chains increases,leading to an increase in the stress concentration value required for the macroscopic failure of the sample.Due to the highest bonding strength,the generation of transgranular cracks in quartz requires a higher concentrated stress value.With increasing volume composition of quartz,the number of transgranular cracks in quartz distributed in the region connecting both loading points increases,which requires many high-strength force chains.The load level rises,leading to an increase in the tensile strength of the numerical sample.

基于GBM的特征选择在心脏病预测中的研究

心脏病患者的人数逐年增多,死亡率居高不下,因此许多学者对基于机器学习的患病人群识别方法进行了研究,但由于心脏病数据的特征过多,构建的模型参数较多,往往导致训练时间过长。因此提出一种利用梯度提升机算法(GBM)对已有的特征选择算法进行改进,在通过减少特征数量实现缩短训练时间的同时,对心脏病患者的预测准确率也得到了提升。首先,采用遗传算法(GA)和基于相关性的特征选择(CFS)进行特征降维,并选出备选特征子集;其次,提出利用GBM实现对模型预测无关特征的筛除;最后,采用支持向量机(SVM)和K近邻算法(KNN)进行准确率预测。实验表明:当预测模型为SVM时,GBM对GA和CFS的综合评价指标的提升分别达到了9.89%和2.2%;当预测模型为KNN时,GBM对GA和CFS的综合评价指标的提升分别达到了7.7%和2.2%,并且GA+GBM在SVM上的表现最好,达到了84.62%,能够有效地提升医疗诊断的效率。

Numerical Study on Aerodynamic Performance of Floating Dual-Rotor Wind Turbines in Heave and Surge Motions

Compared with the traditional wind turbine of a single rotor,dual-rotor wind turbines(DRWTs)have higher wind energy capture efficiency and a more complex structure.Therefore,the aerodynamic performance of the DRWT installed on the floating platform will be greatly affected by the motion caused by wind and wave loads.In this paper,5 MW and 750 kW single rotor wind turbines(SRWTs)are combined into a 5 MW-5 MW DRWT and a 5 MW-750 kW DRWT,and their power output and wake field characteristics in different motions are studied.The flow field is obtained by solving the Reynolds-averaged Navier–Stokes equation(RANS).The overset grid technique is employed to achieve the large-amplitude multiple-degree-of-freedom motion of the DRWT.The overall performance of the 5 MW single rotor wind turbine is determined by a numerical method.For the DRWTs,numerical results show that the surge motion and heave motion both have a negative effect on the power output of the DRWT.The surge motion is a critical factor that causes the power output of the DRWT to periodically change with motion.The average power output of the DRWT influenced by motion is lower than that of a DRWT with a fixed bottom.The surge motion significantly disturbs the wake of the DRWT due to the mutual interference between the upstream and downstream rotors.Under the influence of heave motion,low-velocity regions downstream of the blade tip are enhanced.This study indicates that attenuating the surge and heave motion of offshore DRWT is very significant for improving its efficiency and should be taken into consideration during the design procedure.

基于lightGBM算法对银行账户反欺诈模型的应用

在互联网金融诈骗产业化、规模化且广泛使用各类前沿技术的背景下,以人工智能等为代表的先进技术为金融机构提高反欺诈能力提供了新的方向。为探索机器学习技术在风险管理中的应用,文章以银行账户数据为研究基础,通过有监督训练,基于LightGBM算法的机器学习模型不断学习数据模式,在完成变量选择的同时择优输出针对欺诈风险的概率预测。

基于Light GBM的Facebook学术成果提及动机自动识别

文章研究了基于LightGBM算法的Facebook提及学术成果动机的自动识别,以实现对Facebook提及的大规模深入使用。基于被提及的学术成果、发布学术成果的Facebook用户、Facebook学术成果发布贴子的文本三类背景数据,提取有可能相关的特征,并利用机器学习算法自动识别动机。(1)在这三类背景数据中都发现了与Facebook提及动机明显相关的特征。特别是,相关特征是替代计量关注度得分、合作国家数量、关注者数量、点赞数量、发布学术成果的Facebook用户身份和Facebook帖子的评论数量;(2)LightGBM分类模型对Facebook提及动机的预测精度为0.31。相比之下,没有Facebook帖子文本特征的分类精度为0.35,高于整体特征组合。只有帖子文本特征的分类精度为0.27。结合帖子的长度和语言后,精度提高到0.30;(3)Facebook动机的分类精度与用户的活动有正相关。结合所有特征后,在活跃度方面前25%区间用户的分类精度达到1,第二25%区间的分类精度为0.36,第三25%区间的分类精度为0.32。总之,考虑到Facebook提及动机自动分类的高度复杂性,本研究取得了比较高的分类精度,可以为今后的研究提供参考。

Revealing the Fluttering Motions of Mid-Water Trawl Codend Through Sea Trials:Case Study of Antarctic Krill Trawl Codend

The dynamic coupling between the fluttering motions and hydrodynamic characteristics of codend is essential in understanding the trawl selectivity through fish response and the drag force acting on the whole trawl.This study investigated the effect of towing speed,warp length,warp tension,and catch size on the fluttering motions of Antarctic krill trawl codend during net shooting,towing,and hauling by using sea trial data.The time-periodicity of codend oscillation was analyzed by the Morlet wavelet transform method.Results indicated that the period of codend oscillation was between 50 s and 90 s and showed an increasing trend with the warp tension but a decreased value at the towing stage.The coefficient amplitude of codend oscillation was between 0 and 4 at the net shooting and hauling stages,and between 0.2 and 0.6 at the towing stage.The amplitude of codend oscillation increased with the warp tension,towing speed,and catch size,but decreased with the increase of the warp length.In addition,the period of codend oscillation increased with the towing speed at the net shooting and hauling stages,but decreased at the towing stage.These results from codend fluttering motions can improve the understanding of fish behavior and gear shape that modify the hydrodynamic force on the codend instantaneously.

Effects of the probability of pulse-like ground motions on landslide susceptibility assessment in near-fault areas

Earthquake-induced strong near-fault ground motion is typically accompanied by largevelocity pulse-like component,which causes serious damage to slopes and buildings.Although not all near-fault ground motions contain a pulse-like component,it is important to consider this factor in regional earthquake-induced landslide susceptibility assessment.In the present study,we considered the probability of the observed pulse-like ground motion at each site(PP)in the region of an earthquake as one of the conditioning factors for landslide susceptibility assessment.A subset of the area affected by the 1994Mw6.7 Northridge earthquake in California was examined.To explore and verify the effects of PP on landslide susceptibility assessment,seven models were established,consisting of six identical influencing factors(elevation,slope gradient,aspect,distance to drainage,distance to roads,and geology)and one or two factors characterizing the intensity of the earthquake(distance to fault,peak ground acceleration,peak ground velocity,and PP)in logistic regression analysis.The results showed that the model considering PP performed better in susceptibility assessment,with an area under the receiver operating characteristic curve value of 0.956.Based on the results of relative importance analysis,the contribution of the PP value to earthquakeinduced landslide susceptibility was ranked fourth after the slope gradient,elevation,and lithology.The prediction performance of the model considering the pulse-like effect was better than that reported previously.A logistic regression model that considers the pulse-like effect can be applied in disaster prevention,mitigation,and construction planning in near-fault areas.

Evaluation of underground blast-induced ground motions through nearsurface low-velocity geological layers

Surface ground motion produced by underground blasts is significantly influenced by near-surface geological conditions.However,near-surface low-propagation velocity layers were always ignored in past analyses of ground motions due to their thin thickness.With the rising concern about surface ground motions produced by the ascendant scale and frequentness of underground excavation and mining,close attention is gradually paid to ground blast vibrations.Therefore,systemic experiments were conducted and took seven months in an underground mine to clarify the variation of motion from underground rock to surface ground.The attenuation of surface ground peak particle velocities(PPVs)is compared to that in underground rock,and horizontal amplitudes are compared to vertical amplitudes.Differences between bedrock and surface ground vibrations are analyzed to illustrate the site effect of near-surface lower-propagation velocity layers.One-dimensional site response analysis is employed to quantify the influence of different geological profiles on surface ground vibrations.The experimental data and site response analysis allowed the following conclusions:(1)geological site effects mainly produce decreasing dominant frequency(DF)of surface ground vibrations;(2)the site amplification effect of blast vibration needs to be characterized by peak particle displacement(PPD);(3)shear waves(S-waves)begin to dominate and surface Rayleigh waves(R-waves)develop as blast-induced ground vibrations travel upward through rock and lower-velocity layers to the surface.The comparison of response relative displacement to a critical value is best to assess the potential for cracking on surface structures.

基于BO-LightGBM混合模型的热轧带钢板凸度预测

为了满足日益增长的热轧带钢板凸度预测要求,建立了基于贝叶斯优化(Bayesian Optimization,BO)结合轻量梯度提升机(Light Gradient Boosting Machine,Light GBM)的板凸度预测模型BO-Light GBM。首先通过贝塞尔公式去除异常值,并经过五点三次平滑降噪。其次,建立Light GBM模型进行板凸度预测,最后通过贝叶斯优化算法选择最优参数,得到最优的模型,并与梯度提升决策树(Gradient Boosting Decision Tree,GBDT)、极端梯度增强(Extreme Gradient Boosting,XGBoost)和随机森林(Random Forest,RF)模型XGBoost、GBDT、RF算法进行比较。实验表明,基于BO-light GBM的板凸度预测模型优于XGBoost、GBDT、RF模型,对测试集预测的决定系数(Coefficient of Determination,R2)、平均绝对误差(Mean Absolute Error,MAE)和均方根误差(Root Mean Squared Error,RMSE)分别为0.97、1.49、2.28μm,均优于其他预测模型,且98%以上样本点的预测误差在-4~4μm,满足板凸度预测的精度要求。同时为了进一步验证该模型的预测稳定性,将模型运行100次后,R2、MAE和RMSE的分布仍处于最优位置,所以认为BO-LIght GBM是值得推荐的板凸度预测模型。

基于MIC-LightGBM的综合能源系统多元负荷预测

提出了基于最大互信息系数(Maximal Information Coefficient,MIC)结合轻梯度提升机(Light Gradient Boosting Machine,Light GBM)的综合能源系统多元负荷预测模型。首先利用MIC进行相关因素分析,筛选特征降低预测模型冗余度,最后输入至Light GBM模型获得多元负荷预测结果。经算例验证,基于基学习器集成的Light GBM具有最低预测误差和较快的预测速度。

利用InSAR形变反演滑坡滑面运动特征——以汶川县布瓦村滑坡为例

当前滑坡滑面运动特征获取多采用接触式手段,费时费力、成本高昂,且仅能获取稀疏点位滑面运动状态。因此,本文提出一种非接触式滑坡滑面运动特征获取方法,其基本思路是:基于位错理论建立滑坡表面形变与滑面运动间的关系模型,同时引入滑面运动平滑约束条件,最终以InSAR等技术获取的地表形变为约束,反演获得滑面运动特征。本文选择四川省汶川县布瓦村滑坡作为研究对象,以时序InSAR技术获取的滑坡坡面升降轨形变数据为约束,基于发展的理论方法反演获得布瓦村滑坡滑面几何与运动学参数,并结合野外勘查、Lidar观测以及模型残差较好地验证了反演结果的可靠性,其中反演升降轨模型残差仅3.1 mm·a^(-1)和3.4 mm·a^(-1)。反演结果显示,布瓦村滑坡滑面位于坡表以下平均~15.0 m深度处,滑面整体以沿坡向运动为主,最大运动量级达~80 mm·a^(-1),同时在坡体西侧中下部发现微量的垂直坡向运动。本文提出方法可方便、快速的获取空间连续的滑坡滑面运动特征,为滑坡灾害预警和失稳风险评估提供更为直接可靠的科学参考。

高速履带车辆横摆运动响应特性分析与试验验证

为研究高速履带车辆操纵特性,奠定高速电驱动履带车辆的操纵稳定性评价及控制的基础,建立了高速履带车辆非线性转向动力学模型及2自由度线性转向动力学模型,并对模型进行试验验证。利用2自由度线性模型推导了高速履带车辆横摆运动传递函数,基于此进行了高速履带车辆横摆运动时域和频域响应特性分析,提出履带车辆稳态横摆角速度增益及临界阻尼车速的定义。研究结果表明:履带车辆稳态横摆角速度增益均小于1,履带车辆具有不足转向特性;履带车辆系统阻尼比在1左右;当车速小于临界阻尼车速时,车辆系统为过阻尼系统,横摆角速度响应的上升时间在0.2 s内;当车速等于临界阻尼车速时,车辆系统为临界阻尼系统,横摆角速度响应的上升时间大于10 s;当车速大于临界阻尼车速时,车辆系统为欠阻尼系统,横摆角速度响应的上升时间迅速减小至2~3 s。

基于机器学习算法的心力衰竭10年患病风险可解释预测建模分析

目的建立基于机器学习算法的心力衰竭10年患病风险预测模型,并通过SHAP方法提升模型的可解释性,以提高心力衰竭风险评估的准确性和临床应用价值。方法采用英国生物银行(UK Biobank,UKB)数据库数据,涵盖了40~70岁之间的502349名英国成年人,基于2006~2010年间的基线数据。选取487572例未发生心力衰竭和10374例发生心力衰竭的病例,随访时间为10年,以ICD-10编码定义心力衰竭事件。使用LightGBM、XGBoost、CatBoost三种机器学习算法构建预测模型,在Python和RStudio环境中完成数据预处理、特征选择及模型效能评估,利用SHAP方法可视化解释模型预测结果。结果经过随机欠采样平衡样本后,本研究建立的模型有效预测了10年内心力衰竭的发病情况。LightGBM模型展现出最佳的预测性能,其次是CatBoost和XGBoost。SHAP值分析揭示年龄、胱抑素C、接受治疗或服用药物次数、曾诊断患有心血管疾病、心血管疾病相关多基因风险分数是心力衰竭风险预测的重要影响因素。结论本研究证实了机器学习模型在心力衰竭风险预测中的有效性,特别是LightGBM模型在所有比较的模型中表现最佳。SHAP值的分析为理解模型预测的驱动因素提供了新的视角,有助于临床决策支持和风险管理。

跨断层工程输入地震动模拟及地震响应

为解决跨断层结构输入地震动的问题,揭示其地震响应规律,基于断层物理模型并引入等效脉冲函数,构建考虑地震动空间变化特征的转换矩阵,提出一种高、低频叠加的混合模拟方法,实现断层两侧输入地震动的模拟。首先,依据建立的桥址场地断层模型,采用随机有限断层方法生成目标点位的高频地震动;再由走滑断层两侧地震动的脉冲效应和永久位移的特征,采用不同的等效脉冲模型分别模拟断层平行向和法向的低频脉冲分量,两者在截止频率处采用Butterworth滤波器进行高通和低通滤波,依据场地模型以及走滑断层两侧地震动的空间相干性,建立转换矩阵以模拟其空间变异性,最终将匹配滤波后的高、低频分量在时域叠加得到断层两侧的输入地震动。从时程、频谱以及结构响应三方面验证了模拟结果的合理性。以实际的跨断层悬索桥为研究对象,基于OpenSees建立全桥的三维有限元模型,采用模拟的断层两侧地震动进行动力时程分析。研究结果表明:桥梁跨断层的角度和位置,以及永久位移幅值对跨断层桥梁地震响应有显著影响,较大的残余内力和残余位移是造成跨断层桥梁破坏的重要原因。