【作者】 李敏；

【导师】 陈观林；

【作者基本信息】 浙江大学 ， 计算机技术， 2018， 硕士

【摘要】 随着城市化进程的不断加快,也带来了各种问题和挑战,施工扰民、乱搭乱建、绿地脏乱、油烟污染、无照经营各种等影响街面秩序的案件在城市道路中随处可见,这也是政府相关城管人员日常工作中重点关注的内容。随着物联网技术和数据感知技术的成熟,城市中能获取的数据量迅速增长。利用数据挖掘技术对这些庞杂的数据进行深度挖掘、有效预测,给政府人员日常工作中提供决策指导,是城市道路管理研究的一个重要课题。本文在对宁波市的历史案件数据进行深入研究基础上,提出了一个综合考虑数据长期变化趋势以及短期波动特征的日案件数预测模型。该模型利用具有强非线性处理能力的SVR_BP组合预测模型对数据的非线性特征进行初步拟合,再利用时间序列预测模型对组合预测模型产生的残差进行建模预测,拟合数据不太突出的线性特征,产生的最终预测值能对区域未来的日案件数提供有效预测,从而在城市道路管理中提供相关参考信息。同时,本文在研究云计算技术的基础上,将云计算技术引用到了城市道路管理系统中,其利用HBase实现了海量数据存储,能有效存储急剧积累的数据;引入上述组合日案件数预测模型,为道路管理工作提供日案件数变化趋势等辅助决策信息;利用Impala对历史案件数据实时运算处理,应用数据可视化技术发掘历史数据的内在价值,为管理工作提供信息指导。城市道路管理系统主要面向政府相关人员以及城市居民,包括案件管理模块、当日案件查询模块、案件预测模块以及历史数据统计分析模块等。案件管理模块可供工作人员实时录入、修改、查看道路相关案件;当日案件查询模块以直观的方式提供了各个区的当日已录入案件数及其小时分布图;案件预测模块利用日案件数预测模型为每个区预测未来七天的日案件数;历史数据统计分析模块则基于对历史案件数据的统计分析,以图表形式可视化呈现数据相关特性,如区域月结案率对比图、部门月人均结案数对比图等。更多还原

【Abstract】 With the accelerating process of urbanization,new problems and challenges have also emerged.The problem of construction disturbing,illegal construction in residential area,oil-smoke pollution and unlicensed business activities and so on,which seriously affect street order,can be seen everywhere in urban road.This is also the key concerns in the daily work of government officials.In today’s internet of things environment,using data mining techniques to deeply analyze massive cases data to provide effective guidance is an important topic in the study of management of urban roads.Based on the in-depth study of the historical data about street cases of Ningbo,this paper proposes a prediction model about the num of daily street cases,which takes into account the global and partial characteristics.This model makes use of the SVR_BP combination forecasting model to fit the nonlinear features.Then the arima model is used to catch the linear feature which is hidden in the residual produced by the combined model.The final predict can provide an effective forecast of the number of daily cases in future,thus providing available guidance for the work of urban road management.After intensive study of cloud computing technology,we develop an urban road management system based on the technology.It adopts the distributed database of HBase to design an effective storage scheme to respond to the rapid growth of the scale of data.It also uses the analysis engine of Impala to offer real-time processing and analysis of historical cases and display the result with the data visualization technology.The road management system is mainly for the government and urban residents,including the module of the query of the cases of the day,the prediction of cases and the statistical analysis of historical data.The module of query offers the number of cases having occurred in "current day and its hours distribution.The prediction module makes use of the forecast model to predict the number of daily cases in each district for the next seven days.The analysis module tends to excavate the hidden value behind the massive data.更多还原