不同等级农村公路交通事故严重程度预测研究

为准确分析各因素及其组合对农村公路发生严重交通事故的影响,将道路类型、路侧防护设施、时间段、天气状况、事故位置等11个因素作为自变量,事故严重程度作为因变量,采用改进的Apriori关联算法,找出各影响因素间的内在联系,并求解得到关键因素的组合。然后,分别构建随机森林、梯度提升决策树(Gradient Boosting Decision Tree,GBDT)、极端梯度提升(eXtreme Gradient Boosting,XGBoost)预测模型对农村公路事故严重程度进行预测。结果显示:与改进前相比,改进后的Apriori算法的运行效率和挖掘准确度都有较大提升;相对于随机森林和GBDT模型,XGBoost模型在准确率、召回率、精确率和F1得分等方面表现最优;照明条件、路侧防护设施、道路类型,天气状况是影响农村公路事故严重程度的重要因素,且各影响因素间存在显著的交互效应。

基于集成学习的交通事故严重程度预测研究与应用

目前自动驾驶技术重点是关注如何主动避免碰撞,然而在面对其他交通参与者入侵而导致不可避免的碰撞事故场景时,预测车辆在不同行驶模式下的碰撞严重程度来降低事故严重程度的研究却很少。为此,提出一种双层Stacking事故严重程度预测模型。基于真实交通事故数据集NASS-CDS完成训练,模型输入为车辆传感器可感知得到的事故相关特征,输出为车内乘员最高受伤级别。在第1层中,通过实验对不同学习器组合进行训练,最终综合考虑预测性能以及耗时挑选K近邻、自适应提升树、极度梯度提升树作为基学习器;在第2层中,为降低过拟合,采用逻辑回归作为元学习器。实验结果表明,该方法准确率达到85.01%,在精确率、召回率和F1值方面优于其他个体模型和集成模型,该预测结果可作为智能车辆决策规划模块先验信息,帮助车辆做出正确的决策,减缓事故损害。最后阐述了模型在L_(2)辅助驾驶与L_(4)自动驾驶车辆中的应用,在常规车辆安全防护的基础上进一步提升车辆的安全性。

基于XGBoost模型的交通事故严重程度预测

本文采用美国2022年交通事故数据集,通过SMOTE_NC算法对数据进行过采样,从时空因素、天气因素和路况因素三个方面分析对交通事故的严重程度具有影响的13个因子,通过RStudio软件的XGBoost包里的XGBoost函数建立XGBoost多分类预测模型,验证模型的准确性为89.76%,证明了XGBoost多分类模型的可行性,丰富了交通事故严重程度预测的理论内容,为减少事故影响、保障人民生命财产安全提供保障.

基于集成学习模型的交通事故严重程度时空预测

为探究区域交通事故时空特征,精准预测事故严重程度,给交通运输主管部门提供决策支持,以英国交通事故统计数据作为研究基础,首先,将交通事故时空特征数据转化为网格化数据,并对空间特征进行二维卷积,利用时间特征合并二维卷积为三维卷积,解决网格冲突问题;其次,利用卷积神经网络(CNN)、长短期记忆人工神经网络(LSTM)模型的工作原理建立Stacking模型的基学习器和元学习器;最后,输出结果传入分类与回归树(CART),构建完整的事故严重程度预测集成学习模型。研究结果表明,集成学习模型较单一模型对预测效果更优,其AUC比CNN、LSTM和Conv-LSTM单一模型预测分别提升0.02、0.04和0.01;最终决策树选择中,CART决策树比随机森林(RF)和梯度提升决策树(GBDT)预测效果更优;预测结果在时间纬度上,“严重事故”事件占比较实际低3.95%,在空间纬度上,预测热力区域在0.5~1区间范围内与实际接近。

基于XGBoost的高速公路事故类型及严重程度预测方法

高速公路事故频发,而以往研究未能充分揭示交通流动态特性对事故类型与严重程度的影响。为此研究了基于动态交通流数据的高速公路事故类型与严重程度的预测方法。从高速公路门架数据中提取流量、密度、速度等交通流数据,同时考虑时间特征以及时间和空间不均匀性特征的数据,与事故数据相匹配构成全样本。建立了基于极端梯度提升树(extrem Gradient Boosting,XGBoost)算法的预测模型,预测事故是否发生、事故类型以及事故严重程度。分别考虑追尾事故和其他事故2种事故类型、有人员伤亡和仅财产损失2种事故严重程度,模型的结果表明:(1)上下游速度差大、低速、路段车流量大且频繁分流、合流条件下交通事故风险较高;(2)低速、路段车辆多且合流、分流交通量大、上下游速度差大的情况下发生追尾事故的风险更高;(3)路段车流量较少且追尾事故发生于周末或夜间可能会增大事故严重程度。将常用机器学习算法与XGBoost算法的预测效果进行对比,XGBoost事故类型预测模型与事故严重程度预测模型的ROC曲线下面积(Area Under Curve,AUC)分别达到了0.76和0.88——相比于序列Logistic、高斯朴素贝叶斯、线性SVM、随机森林以及神经网络等其他常用算法,平均分别提升了0.08和0.24。这表明基于XGBoost建立的模型具有较好的预测性能。研究结果为高速公路路段实时交通流状态预警提供了可靠手段,进而可以提升高速公路行车安全。

急性胰腺炎严重程度预测

急性胰腺炎是多种病因导致胰酶在胰腺内被激活后引起胰腺组织自身消化、水肿、出血甚至坏死的炎症反应。临床以急性上腹痛、恶心、呕吐、发热和血胰酶增高等为特点。病变程度轻重不等,轻者以胰腺水肿为主,临床多见,病情常呈自限性,愈后良好,又称为轻症急性胰腺炎。少数重症患者的胰腺出血坏死,常继发感染、腹膜炎和休克等,病死率高,称为重症急性胰腺炎。随着该疾病临床上的发展出现了各种对急性胰腺炎严重程度的预测因子,本文从急性胰腺炎患者因素、免疫因素、细胞因子因素、生化因素、影像因素、评分系统、对目前临床上广泛应用且研究的预测因子进行总结。

考虑交通事故时空效应的高速公路交通事故严重程度预测

为提高高速公路交通事故处置效率,降低交通事故造成的生命财产损失,提出一种基于卷积神经网络(CNN)的高速公路交通事故严重程度预测方法。首先确定交通事故严重程度的影响因素,并根据交通事故损失情况和对事故路段的影响将事故划分为轻微事故、一般事故和严重事故三类;然后采用梯度提升决策树法计算各影响因素对交通事故严重程度影响的权重等级,根据权重等级将影响因素特征矩阵转化为灰度图像,即将交通事故数据的单一特征关系转换为包含并行组合关系的方形特征矩阵;以事故数据的时空特性作为输入,构建基于CNN的高速公路交通事故严重程度预测模型,并对模型进行评估与检验。结果表明,在高速公路交通安全事故严重程度评估中,相较于随机森林机器模型和逻辑回归模型,文中模型的识别率明显提高,对严重交通事故的测试准确率超过80%。

青藏高原双车道事故严重程度预测模型的建立

分析了青藏高原地区的交通事故数据,发现大部分交通事故都是死伤人数较高的事故,即死伤事故占比较大。将事故严重程度作为因变量,选取海拔、氧含量、天气、事故地点的线形、事故地点的环境、肇事车辆的类型、交通量、大车比例、大小车的速度差等9个因素作为自变量,建立了累积Logistic回归预测模型,研究了影响高原地区交通事故严重程度的因素。结果表明:海拔、大车比例、交通量、肇事车辆类型和大小车的速度差与交通事故严重程度有显著的相关性。预测模型的建立能够为相关公路管理部门制定安全措施提供理论依据。

ORESP:基于有序回归的软件缺陷严重程度预测方法

为提高软件缺陷严重程度的预测性能,通过充分考虑软件缺陷严重程度标签间的次序性,提出一种基于有序回归的软件缺陷严重程度预测方法ORESP。该方法首先使用基于Spearman的特征选择方法来识别并移除数据集内的冗余特征,随后使用基于比例优势模型的神经网络来构建预测模型。通过与五种经典分类方法的比较,所提的ORESP方法在四种不同类型的度量下均可取得更高的预测性能,其中基于平均0-1误差(MZE)评测指标,预测模型性能最大可提升10.3%;基于平均绝对误差(MAE)评测指标,预测模型性能最大可提升12.3%。除此之外,发现使用基于Spearman的特征选择方法可以有效提升ORESP方法的预测性能。

一种变粒度缺陷报告严重程度预测方法

已有工作一般将缺陷报告严重程度预测问题建模为粗粒度分类问题(即二分类问题)或细粒度分类问题(即多分类问题),更细的预测粒度意味着对分类准确性提出更高的要求,因而预测性能通常会下降。通过有效利用粗粒度分类的高预测性能与细粒度分类更为精细的预测粒度,提出一种变粒度缺陷报告严重程度预测(variable granularity bug report severity prediction,VG-BSP)方法,VG-BSP方法首先对缺陷报告进行基于二分类的粗粒度划分,然后根据二分类的结果,选择对应的预测模型以预测出具体的严重程度(即基于多分类的细粒度划分)。为了验证VG-BSP方法的有效性,搜集了来自Eclipse的两个子项目JDT和CDT的缺陷报告和来自Mozilla项目的缺陷报告;并将VG-BSP方法与逻辑回归(logistic regression,LR)方法、K近邻(K-nearest neighbors,KNN)方法、决策树(decision tree,DT)方法、随机森林(random forest,RF)方法、支持向量机(support vector machine,SVM)方法等经典细粒度分类方法进行了对比。实验结果表明:基于macro-F1指标,VG-BSP方法的性能与已有的细粒度方法相比平均可提升19.6%。除此以外,使用LR分类算法和Spacy词向量可以使得VG-BSP方法取得最好的性能。

基于一种组合新模型的翻车事故严重性预测

由于近年来发生在高速公路上的翻车事故较多,且预测翻车事故严重性的文献较少,故通过一种组合新模型来预测翻车事故的严重程度。通过收集的2012年—2016年1 939起高速公路翻车事故数据,采用Logistic回归与朴素贝叶斯组合的新模型作为翻车事故严重性的预测方法。为了进一步验证组合预测模型的优越性,构建了朴素贝叶斯预测模型和Logistic回归预测模型。对比这3个预测模型的预测数据正确率值,表明组合预测模型效果更好。同时,11个候选变量经过单因素分析和多因素分析,选出7个显著性变量:驾驶员年龄、事故涉及车型、事故涉及车辆数、事故发生时段、天气条件、事故是否发生在匝道、肇事车辆是否超载超限。当相应道路或环境条件得到改善后,组合预测模型有助于道路管理者了解翻车事故严重程度可能的变化情况;当出现一起翻车事故时,可根据已知的有限信息利用组合预测模型预估事故严重程度,以便开展后期救援。

急性心肌梗死临床诊断中肌钙蛋白I结合心肌酶谱检验的作用

本研究旨在探究在急性心肌梗死临床诊断中,肌钙蛋白I结合心肌酶谱检验的作用。方法 回顾性分析2021年3月至2023年3月在柳州市中医医院就诊的80例急性心肌梗死患者,选取同期80例健康体检者作为对照组。对比观察两组肌钙蛋白I的浓度、观察组不同时期肌钙蛋白I的浓度、观察组治疗前后心肌酶谱、肌钙蛋白I的浓度。结果 观察组肌钙蛋白I的浓度显著高于对照组(p<0.05)、观察组肌钙蛋白I的浓度在心肌损伤后,cTnI的水平会在数小时内上升、观察组治疗后肌钙蛋白I的浓度显著下降(p<0.05)。结论 本研究表明,肌钙蛋白I在急性心肌梗死的临床诊断中具有重要作用。其高度的敏感性和特异性,使其成为急性心肌梗死诊断的关键辅助工具。然而,在应用过程中,仍需综合考虑其他临床指标以确保准确诊断和有效治疗。

Value of adipokines in predicting the severity of acute pancreatitis:Comprehensive review

AIM:To analyze the prognostic value of adipokines in predicting the course,complications and fatal outcome of acute pancreatitis(AP).METHODS:We performed the search of PubMed database and the systemic analysis of the literature for both experimental and human studies on prognostic value of adipokines in AP for period 2002-2012.Only the papers that described the use of adipokines for prediction of severity and/or complications of AP were selected for further analysis.Each article had to contain information about the levels of measured adipokines,diagnosis and verification of AP,to specify presence of pancreatic necrosis,organ dysfunction and/or mortality rates.From the very beginning,study was carried out adhering to the PRISMA checklist and flowchart for systemic reviews.To assess quality of all included human studies,the Quality Assessment of Diagnostic Accuracy Studies tool was used.Because of the high heterogeneity between the studies,it was decided to refrain from the statistical processing or meta-analysis of the available data.RESULTS:Nine human and three experimental studies were included into review.In experimental studies significant differences between leptin concentrations at 24 and 48 h in control,acute edematous and acute necrotizing pancreatitis groups were found(P = 0.027 and P < 0.001).In human studies significant differences between leptin and resitin concentrations in control and acute pancreatitis groups were found.1-3 d serum adiponectin threshold of 4.5 μg/mL correctly classified the severity of 81% of patients with AP.This threshold yielded a sensitivity of 70%,specificity 85%,positive predictive value 64%,negative predictive value88%(area under curve 0.75).Resistin and visfatin concentrations differ significantly between mild and severe acute pancreatitis groups,they correlate with severity of disease,need for interventions and outcome.Both adipokines are good markers for parapancreatic necrosis and the cut-off values of 11.9 ng/mL and 1.8 ng/mL respectively predict the high ranges of radiological scores.However,the review revealed that all nine human studies with adipokines are very different in terms of methodology and objectives,so it is difficult to generalize their results.It seems that concentrations of the leptin and resistin increases significantly in patients with acute pancreatitis compared with controls.Serum levels of adiponectin,visfatin and especially resitin(positive correlation with Acute Physiology and Chronic Health Evaluation Ⅱ,Ranson and C-reactive protein) are significantly different in mild acute pancreatitis and severe acute pancreatitis patients,so,they can serve as a markers for the disease severity prediction.Resistin and visfatin can also be used for pancreatic and parapancreatic necrosis prediction,interventions needs and possible,outcome.CONCLUSION:High levels of adipokines could allow for prediction of a severe disease course and outcome even in small pancreatic lesions on computed tomography scans.

Is the ratio of apoB/apoA-1 the best predictor for the severity of coronary artery lesions in Chinese diabetics with stable angina pectoris？ An assessment based on Gensini scores

Background There is a paucity of data about the best lipid ratio predicting the severity of coronary artery disease （CAD） in patients with diabetes mellitus. We determined the relationship between five conventional lipid ratios and the extent of coronary artery lesions in Chinese Type 2 diabetics with stable angina pectoris （SAP）. Methods A prospective cohort study within 373 type 2 diabetic patients diag- nosed with stable CAD by coronary angiography was performed. All patients were classified into three groups according to the tertiles of Gensini scores （GS, low group 〈 8 points n = 143; intermediate group 8-28 points, n = 109; high group 〉 28 points, n = 121）. Association between the ratios of apolipoprotein （apo） B and apoA-1, total cholesterol and high density lipoprotein cholesterol （TC/HDL-C）, triglycerides and HDL-C （TG/HDL-C）, low density lipoprotein cholesterol and HDL-C （LDL-C/HDL-C）, Non-HDL-C/HDL-C and GS were evaluated using the receivers operating characteristic （ROC） curves and multivariate logistic regression models. Results The ratio of apoB/apoA-1, TC/HDL-C, LDL-C/HDL-C, and Non-HDL-C/HDL-C were correlated with Gensini scores. Area under the ROC curves for predicting high Gensini scores in the ratios of apoB/apoA-1, TC/HDL-C, LDL-C/HDL-C and Non-HDL-C/HDL-C were 0.62, 0.60, 0.59 and 0.60, respec- tively （P 〈 0.005 for all）. According to multivariate logistic regression analysis after adjusted with demographic characteristic and other lipid parameters, the ratio of apoB/apoA-1 is qualified as an independent discriminator for the severity of CAD. However, after fiu-ther adjusting different baseline variables, such as left ventricular ejective fraction, hemoglobin Alc, leukocytes count and serum creatinine, none of the above lipid ratios remained. Conclusions Compared with other lipid parameters, the ratio of apoB/apoA-1 appears to be more significantly correlated with the extent of coronary artery lesions in Chinese diabetics, but it was not an independent predictor in these settings.

Could quantitative liver function tests gain wide acceptance among hepatologists?

It has been emphasized that the assessment of residual liver function is of paramount importance to determine the following： severity of acute or chronic liver diseases independent of etiology; long-term prognosis; step-bystep disease progression; surgical risk; and efficacy of antiviral treatment. The most frequently used tools are the galactose elimination capacity to asses hepatocyte cytosol activity, plasma clearance of indocyanine green to assess excretory function, and antipyrine clearance to estimate microsomal activity. However, a widely accepted liver test （not necessarily a laboratory one） to assess quantitative functional hepatic reserve still needs to be established, although there have been various proposals. Furthermore, who are the operators that should order these tests？ Advances in analytic methods are expected to allow quantitative liver function tests to be used in clinical practice.

Chinese College Students Seeking Psychological Help From Professionals^Rationally Should but Emotionally not？

The purpose of the present research was to investigate the predictors to Chinese college students seeking psychological help from professionals. By surveying 1,408 Chinese college students at five universities in China's Mainland, the results showed that among the factors examined, problem severity and help-seeking attitudes directly predicted help-seeking intention, while gender and subjective norms had a significant effect on college students＇ professional psychological help-seeking intention through help-seeking attitudes. Overall, the model explained 25% and 38.0% （for half-1 and half-2 data sets, respectively） of the variances of help-seeking intention. The results indicated that mainland Chinese college students rationally knew that they should choose to seek such professional help when the problem got severer enough, although they were emotionally reluctant to seek professional psychological help due to negative subjective norms around them. The implications and limitations were discussed.

基于BiLSTM的突发性交通事故严重程度预测研究

突发性交通事故严重程度预测是道路交通安全管理的研究热点之一。采用省域交通事故记录为数据来源,选取时段、天气、受伤人数等14个影响因素作为自变量,提出一种基于深度学习的突发性交通事故严重程度预测方法。依据事故持续时间将突发性交通事故划分为轻微、一般和严重三个等级,建立基于双向长短时记忆网络(BiLSTM)的突发性交通事故严重程度预测模型。结果表明,BiLSTM模型的F1值最高,优于其他模型。

厦门市重症恙虫病的危险因素分析

目的调查重症恙虫病的危险因素,为预测其严重程度和临床决策提供依据。方法收集厦门市2012—2015年6家医疗机构报告的恙虫病病例,对病例基本情况和实验室报告等资料,进行回顾性分析。结果厦门市2012—2015年共报告恙虫病264例,其中重症64例(24.2%)。多因素logistic回归分析表明:并发弥散性血管内凝血(DIC)、淋巴细胞≥40%、血小板计数<100×109/L及白蛋白<3.0g/dL,是重症恙虫病的危险因素。结论重症监护和积极治疗恙虫病并发症,可降低重症恙虫病病死率。

Empirical analysis of network measures for predicting high severity software faults

Network measures are useful for predicting fault-prone modules. However, existing work has not distinguished faults according to their severity. In practice, high severity faults cause serious problems and require further attention. In this study, we explored the utility of network measures in high severity faultproneness prediction. We constructed software source code networks for four open-source projects by extracting the dependencies between modules. We then used univariate logistic regression to investigate the associations between each network measure and fault-proneness at a high severity level. We built multivariate prediction models to examine their explanatory ability for fault-proneness, as well as evaluated their predictive effectiveness compared to code metrics under forward-release and cross-project predictions. The results revealed the following:(1) most network measures are significantly related to high severity fault-proneness;(2) network measures generally have comparable explanatory abilities and predictive powers to those of code metrics; and(3) network measures are very unstable for cross-project predictions. These results indicate that network measures are of practical value in high severity fault-proneness prediction.