A hybrid deep learning based traffic flow prediction method and its understanding

Deep neural networks (DNNs) have recently demonstrated the capability to predict traffic flow with big data. While existing DNN models can provide better performance than shallow models, it is still an open issue of making full use of spatial-temporal characteristics of the traffic flow to improve their performance. In addition, our understanding of them on traffic data remains limited. This paper proposes a DNN based traffic flow prediction model (DNN-BTF) to improve the prediction accuracy. The DNN-BTF model makes full use of weekly/daily periodicity and spatial-temporal characteristics of traffic flow. Inspired by recent work in machine learning, an attention based model was introduced that automatically learns to determine the importance of past traffic flow. The convolutional neural network was also used to mine the spatial features and the recurrent neural network to mine the temporal features of traffic flow. We also showed through visualization how DNN-BTF model understands traffic flow data and presents a challenge to conventional thinking about neural networks in the transportation field that neural networks is purely a “black-box” model. Data from open-access database PeMS was used to validate the proposed DNN-BTF model on a long-term horizon prediction task. Experimental results demonstrated that our method outperforms the state-of-the-art approaches.     

Predictive power of inspection outcomes for future shipping accidents – an empirical appraisal with special attention for human factor aspects

This paper investigates whether deficiencies detected during port state control (PSC) inspections have predictive power for future accident risk, in addition to other vessel-specific risk factors like ship type, age, size, flag, and owner. The empirical analysis links accidents to past inspection outcomes and is based on data from all around the globe of PSC regimes using harmonized deficiency codes. These codes are aggregated into eight groups related to human factor aspects like crew qualifications, working and living conditions, and fatigue and safety management. This information is integrated by principal components into a single overall deficiency index, which is related to future accident risk by means of logit models. The factor by which accident risk increases for vessels with above average compared to below average deficiency scores is about 6 for total loss, 2 for very serious, 1.5 for serious, and 1.3 for less-serious accidents. Relations between deficiency scores and accident risk are presented in graphical format. The results may be of interest to PSC authorities for targeting inspection areas, to maritime administrations for improving asset allocation based on prediction scenarios connected with vessel traffic data, and to maritime insurers for refining their premium strategies.     

汶川地震震中区高速公路隧道超前地质预报方法

映汶高速公路公路位于四川汶川大地震震中,受地震强烈构造作用影响,隧址区岩体裂隙发育,松散破碎,存在构造富水,地震垮塌堆积体众多,地质条件极为复杂。文章根据本工程隧道的工程地质问题,通过对国内外隧道超前地质预报常用方法的分析,提出了针对本工程的隧道超前地质预报方案,并对地质预报的实施应用及效果进行了举例说明。     

提高TSP探测精度的途径

TSP作为目前最先进的隧道地质超前预报探测仪器，得到了广泛的应用。但是由于在现实中存在各种问题，从而导致该仪器的预测精度受到了极大限制，主要阐述如何提高其预测精度，更好地为隧道的建设服务。     

水下双层圆柱壳辐射声场欠定分离评估方法研究

水下双层加筋圆柱壳振动和辐射声场的评估对其辐射噪声监测和控制具有重要工程意义。文中通过结构振动模态参与因子向量自身的稀疏特性,分析提出了一种基于结构振动的辐射噪声欠定分离评估方法,可实现有限振动测点情况下的水下复杂结构振动和辐射声场的有效评估。数值和试验结果验证了文中方法的有效性,且所需要的振动测点数目少,具有良好的工程适用性。     

Multi-step prediction of experienced travel times using agent-based modeling

This paper develops an agent-based modeling approach to predict multi-step ahead experienced travel times using real-time and historical spatiotemporal traffic data. At the microscopic level, each agent represents an expert in a decision-making system. Each expert predicts the travel time for each time interval according to experiences from a historical dataset. A set of agent interactions is developed to preserve agents that correspond to traffic patterns similar to the real-time measurements and replace invalid agents or agents associated with negligible weights with new agents. Consequently, the aggregation of each agent’s recommendation (predicted travel time with associated weight) provides a macroscopic level of output, namely the predicted travel time distribution. Probe vehicle data from a 95-mile freeway stretch along I-64 and I-264 are used to test different predictors. The results show that the agent-based modeling approach produces the least prediction error compared to other state-of-the-practice and state-of-the-art methods (instantaneous travel time, historical average and k-nearest neighbor), and maintains less than a 9% prediction error for trip departures up to 60 min into the future for a two-hour trip. Moreover, the confidence boundaries of the predicted travel times demonstrate that the proposed approach also provides high accuracy in predicting travel time confidence intervals. Finally, the proposed approach does not require offline training thus making it easily transferable to other locations and the fast algorithm computation allows the proposed approach to be implemented in real-time applications in Traffic Management Centers.     

基于LSTM的舰船运动姿态短期预测

舰船的六自由度运动状态形成复杂的非线性过程,运动姿态会受到耦合作用、不定周期、噪声信号以及混沌特性等因素的干扰,因此很难得到精确的预测结果.为了提升舰船运动姿态的预测精度,利用舰船时间序列的特点,建立了基于长短期记忆单元(LSTM)模型,对其进行了舰船姿态预测仿真,将结果与时间序列分析法的结果进行对比.实例分析表明:基于LSTM模型的预测方法具有精确度高、易实现的特点.这为舰船运动短期预测提供了一个新的思路和方法.     

突涌地质隧道TSP法弹性参数判识研究

结合云南4座在建高风险隧道典型突涌实例,对不同不良地质发生突涌的地质情况及突涌段落地震波反射法(TSP)的预报成果进行总结,分析各类突涌发生的工程地质条件和地震波反射法物理参数,归纳隧道突涌地质特征,得出大规模突涌时地震波反射法物理参数中纵波速度V_p、横波波速V_s、纵横波速比V_p/V_s和泊松比σ四项参数的判断标准:(1)富水时,V_p/V_s与σ都呈增长趋势,V_p/V_s变化率增长约5%以上,σ变化率增长约10%以上;(2)富水时,V_p/V_s由1.7→2.0变化,σ由0.25→0.3变化;(3)层状裂隙或构造破碎时,V_p均呈下降趋势;(4)存在裂隙及富水通道时,V_s均呈下降趋势;(5)V_p和V_s同时下降时,围岩破碎及地下水发育程度同步上升;(6)V_p上升、V_s下降时,围岩完整程度变化不大,地下水或裂隙发育程度上升。     

基于改进极限学习机的公交站点 短时客流预测方法

以公交车IC 卡和GPS数据为基础，提出了一种基于改进粒子群算法优化极限学习机(IPSO-ELM)的公交站点短时客流预测模型.依托IC 卡和GPS 数据在站点的特征表现和内在联系，定义了站点间距，并分析了站间距和车辆到总站距离间的联系；提出了公交乘客上车站点确定方法，进而得到公交站点上车客流量；通过分析公交客流数据特征，确定ELM输入参数维度，并采用IPSO 算法找到ELM的最优隐含层节点参数；最后依托广州市19 路公交车客流数据仓库进行了方法验证.结果表明：所用优化后的ELM方法预测误差在10%以内，并与应用广泛的SVM、ARIMA和传统ELM模型进行对比分析，发现改进的ELM方法拥有更高的可靠性和泛化性能.     

考虑前序路段状态的公交到站时间双层BPNN 预测模型

为提高公交到站时间预测精度，提出基于双层BPNN与前序路段状态的综合预测模型. 基于静态变量及顶层BPNN模型预测车辆到达每个站点的初始行程时间，利用K-means 聚类及马尔科夫链模型基于前序路段状态预测目标路段行驶时间；将上述两个模型的预测值及上一班次车辆的行程时间作为输入变量，基于底层BPNN模型预测车辆在目标路段的行程时间，进而动态调整车辆到达每个站点的时间. 以上海市791 路公交车早晚高峰各路段的行程时间为例进行模型测试，并与其他4 种模型进行比较. 结果表明，所提模型具有较高的预测精度，尤其在雨天，比传统BPNN模型预测精度提高57.25%.