Exploring the uncertainty of activity zone detection using digital footprints with multi-scaled DBSCAN

ABSTRACT

The density-based spatial clustering of applications with noise (DBSCAN) method is often used to identify individual activity clusters (i.e., zones) using digital footprints captured from social networks. However, DBSCAN is sensitive to the two parameters, eps and minpts. This paper introduces an improved density-based clustering algorithm, Multi-Scaled DBSCAN (M-DBSCAN), to mitigate the detection uncertainty of clusters produced by DBSCAN at different scales of density and cluster size. M-DBSCAN iteratively calibrates suitable local eps and minpts values instead of using one global parameter setting as DBSCAN for detecting clusters of varying densities, and proves to be effective for detecting potential activity zones. Besides, M-DBSCAN can significantly reduce the noise ratio by identifying all points capturing the activities performed in each zone. Using the historic geo-tagged tweets of users in Washington, D.C. and in Madison, Wisconsin, the results reveal that: 1) M-DBSCAN can capture dispersed clusters with low density of points, and therefore detecting more activity zones for each user; 2) A value of 40 m or higher should be used for eps to reduce the possibility of collapsing distinctive activity zones; and 3) A value between 200 and 300 m is recommended for eps while using DBSCAN for detecting activity zones.     

Application of tank,NAM, ARMA and neural network models to flood forecasting

Two lumped conceptual hydrological models, namely tank and NAM and a neural network model are applied to flood forecasting in two river basins in Thailand, the Wichianburi on the Pasak River and the Tha Wang Pha on the Nan River using the flood forecasting procedure developed in this study. The tank and NAM models were calibrated and verified and found to give similar results. The results were found to improve significantly by coupling stochastic and deterministic models (tank and NAM) for updating forecast output. The neural network (NN) model was compared with the tank and NAM models. The NN model does not require knowledge of catchment characteristics and internal hydrological processes. The training process or calibration is relatively simple and less time consuming compared with the extensive calibration effort required by the tank and NAM models. The NN model gives good forecasts based on available rainfall, evaporation and runoff data. The black‐box nature of the NN model and the need for selecting parameters based on trial and error or rule‐of‐thumb, however, characterizes its inherent weakness. The performance of the three models was evaluated statistically. Copyright © 2000 John Wiley & Sons, Ltd.     

Virtual geotechnical laboratory experiments using a simulator

The details of a test simulator that provides a realistic environment for performing virtual laboratory experimentals in soil mechanics is presented. A computer program ‘Geo‐Sim’ that can be used to perform virtual experiments, and allow for real‐time observations of material response is presented. The results of experiments, for a given set of input parameters, are obtained with the test simulator using well‐trained artificial neural‐network‐based soil models for different soil types and stress paths. Multimedia capabilities are integrated in Geo‐Sim, using software that links and controls a laser disc player with a real‐time parallel processing ability. During the simulation of a virtual experiment, relevant portions of the video image of a previously recorded test on an actual soil specimen are dispalyed along with the graphical presentation of response from the feedforward ANN model predictions. The pilot simulator developed to date includes all aspects related to performing a triaxial test on cohesionless soil under undrained and drained conditions. The benefits of the test simulator are also presented. Copyright © 2000 John Wiley & Sons, Ltd.     

Modelling of shearing behaviour of a residual soil with Recurrent Neural Network

Modelling of shear behaviour of residual soils is difficult in that there is a significant variability in constituents and structures of the soil. A Recurrent Neural Network (RNN) is developed for modelling shear behaviour of the residual soil. The RNN model appears very effective in modelling complex soil shear behaviour, due to its feedback connections from an hidden layer to an input layer. Two architectures of the RNN model are designed for training different sets of experimental data which include strain-controlled undrained tests and stress-controlled drained tests performed on a residual Hawaiian volcanic soil. A dynamic gradient descent learning algorithm is used to train the network. By training only part of the experimental data the network establishes neural connections between stress and strain relations. Although the soil exhibited significant variations in terms of shearing behaviour, the RNN model displays a strong capability in capturing these variabilities. Both softening and hardening characteristics of the soil are well represented by the RNN model. Isotropic and anisotropic consolidation conditions are precisely reflected by the RNN model. In undrained tests, pore water pressure responses at various loading stages are simultaneously simulated. With a RNN model designed for a special drained test, the network is able to capture abrupt changes in axial and volumetric strains during shearing courses. These good agreements between the measured data and the modelling results demonstrate the desired capability of the RNN model in representing a soil behaviour. © 1998 John Wiley & Sons, Ltd.     

基于超前钻探测试的隧道地层智能识别方法

可靠地识别掌子面前方地层是保证隧道工程稳定与安全的重要因素之一。传统的超前地质预报方法不能同时保证有高识别精度、低实施成本和占用少的施工时间,对于不同地质情况的地层识别通用性不强。在传统超前钻孔的同时获取掌子面前方围岩钻探测试数据,实时获取不同深度岩层情况,将大大提高超前预报效率,方便快捷,不影响施工,但目前缺乏客观、准确的地层识别方法。提出了一种基于神经网络的钻探测试数据智能分析和地层识别方法,对楚大高速公路九顶山隧道超前钻探测试数据进行了深入分析,通过隧道开挖后所揭示地层对分析方法进行了验证。结果表明：单一钻进参数用于地层识别的错误率在35%左右,打击能和打击数、送水压力和送水流量的参数组合不能显著提升地层识别准确率;钻进速度、扭矩、回转数、推进力的参数组合可降低地层识别错误率至22%。在神经网络模型中引入钻进参数的标准差,可大幅降低错误率,可使地层划分错误率下降9%～12%;多参数组合下的神经网络钻探测试神经网络模型对随机抽样的地层识别错误率小于10%,对单个钻孔的地层识别错误率小于14%。     

内蒙古锡林郭勒盟牧区雪灾风险评估研究

选取内蒙古锡林郭勒盟与积雪有关的雪灾致灾指标, 以气温、 风速为气象条件孕灾环境指标, 坡度、 植被盖度为下垫面孕灾环境指标, 人口密度、 牧民纯收入、 人均GDP、 牲畜超载率等数据为承灾体脆弱性指标, 基于BP方法、 层次分析法、 建立了内蒙古锡林郭勒盟白灾综合风险评价体系, 并对其进行了风险评价与区划。为提高灾害评估的准确率, 白灾的灾害等级是以月为尺度进行评定, 选取的气象指标多数都是以月为尺度的指标。研究表明： 白灾与积雪因子高度相关, 是气候灾害, 积雪、 低温、 大风等气象因子长期作用的结果。对白灾尝试用BP神经网络法进行风险评估, 评估的灾害等级和实际灾害等级十分吻合, 用训练好的神经网络对各个旗县（1980 - 2015年）的白灾进行了风险评估, 评估效果理想。因此, 可以通过数值预报产品、 气候预测产品获取相关评价因子, 采用BP方法形成白灾风险预评估产品, 进而应用于雪灾风险评估业务中, 为相关部门提供决策依据。     

基于BP神经网络和FEFLOW模型模拟预测多年冻土活动层温度

土壤温度是陆面过程中地-气系统间能量与物质交换的重要参数, 它的动态变化及其对气候变化的响应也是研究陆面过程的关键问题之一。在全球变暖背景下, 研究青藏高原多年冻土活动层土壤热状况动态变化, 对深入了解高原活动层厚度的变化特征及下垫面的热力作用均有重要意义。利用BP神经网络模型, 对青藏高原风火山地区的地表温度进行了模拟, 并利用输出的地表温度驱动FEFLOW模型对研究区活动层不同深度土壤温度进行了模拟。与各深度土壤温度观测值对比发现, 均方根误差介于0.09 ~ 1.78 ℃, 纳什效率系数介于0.86 ~ 0.98, 模拟效果良好。结合BP神经网络模型和FEFLOW模型预测了研究区未来50年活动层热状况的动态变化过程, 结果表明： 在0.02、 0.048、 0.07 ℃·a^-1三种升温情景下, 50年后研究区活动层厚度将分别增加19.4、 51.8、 64.7 cm, 土壤升温幅度随着深度的增加逐渐减小。同时发现, 随着气温不同程度的升高, 土壤开始融化的时间在不断提前, 开始冻结的时间则不断延迟, 这种规律随着土壤深度的增加而减弱, 但不同深度土壤冻融过程对气温升高的响应差异却随着增温速率的增大而逐渐减小。     

基于Siamese卷积神经网络的影像瓦片变化检测技术

针对地理信息变化较快而传统更新方式效率不高的问题,目前许多学者提出了各类变化检测的方法,但这些方法大都是基于影像数据进行试验,对影像预处理要求较高,且检测精度的稳定性较差,受数据源质量影响较大。而天地图、百度地图、谷歌地图等地图中均可免费下载各种级别的影像瓦片,因此本文提出利用天地图影像瓦片进行试验,采用Siamese卷积神经网络（SCNN）和深度学习技术,开发基于SCNN的高精度变化监测算法,以快速发现变化区域,实现地理信息变化信息检测。     

城市交通热点区域识别及空间交互分析

出租车载客数据可以用于研究居民的出行特征,提取城市的交通热点区域,但对城市交通热点区域的交互关系研究相对较少。本文以纽约市的出租车行程记录数据为数据源,利用交通小区划分结合出租车载客数据提取城市交通热点区域,基于复杂网络的方法对不同日期类型和天气情况的城市交通热点区域空间交互网络进行研究并进行社区发现。结果表明,热点区域受城市核心区的影响而聚集在核心区域周围,城市内社区的形成可以克服地形和行政区域等因素的影响。研究结论有望为城市规划、城市交通管理、出租车调度、以及人们的出行等提供信息参考。     

基于BP人工神经网络平潭海域赤潮叶绿素a浓度模型演算研究

以福建平潭海域为研究对象、以叶绿素a浓度为输出指标,根据2009-2018年赤潮期数据规律及2013-2017年海洋监测数据主成分分析结果,对拟构建的BP模型进行输入指标筛选,选定结果包括4个气象因子和4个水质因子。基于此结果,对2013-2017年的698组海洋监测数据中叶绿素a浓度进行归一化处理并进行模型演算,随机选取80%数据作为演算模型的训练样本,其余进行模型验证。通过交叉变换输入指标,寻求最优的输入节点组合,以气温、溶解氧浓度、日照时长指标为输入参数时,BP模型误差较小(均方根误差为0.05μg/L,平均绝对误差为0.03μg/L),演算结果精度较高(可决系数R~2=0.81)。以上结果表明,气温、溶解氧浓度和日照时长对叶绿素a浓度表征效果较好,可为平潭海域以叶绿素a浓度作为判定指标的赤潮预警研究提供参考。