Urban Grid Management Incidents Pattern Mining And Prediction

The rapid development of urbanization has brought great challenges to the management of cities.The management and early warning of urban incidents have become an important part of urban sustainable development.In order to achieve a good governance effect,city management faces many challenges,such as how to find city incidents efficiently,how to deal with the incident effectively.How to warn escalation of the incident in an early stage and how to accurately forecast the temporal and spatial distribution of the incidents have become the most important one.Shanghai Grid Center is an important part of smart city management in Shanghai.It is mainly responsible to collect data,to arrange and coordinate the disposal of the incidents,to count and to analyze the data of management incidents.It divides the city into small grids according to specific criteria.There are staff in charge of each grid and report incidents when incidents arise.Based on the real business data and urgent requirements of the Grid Center in Xuhui District and Pudong District of Shanghai.This research mainly studies the following contents.It collects the corresponding data from the grid management system and the basic information of facilities to establish a multi-level data warehouse for multi-source data association and integration.It also does clustering analysis of city management incidents.When an incident occurs somewhere in the city,the corresponding Grid Center staff and citizens will report this incident to the Grid Center.Due to the level of severity and due to the speed of disposal,there will be multiple reports to a same incident.To cluster the same incidents can help the management keep track of the information such as the severity,the evolvement and the disposal of this incident.In this paper,an innovative method of event clustering for grid management data is proposed,which uses the similarity of event occurrence time,the similarity of occurrence geographical location and the semantic similarity of event description to build the measure of similarity of events.Then,DBSCAN density clustering method is used to achieve event clustering.By tuning the parameters of this algorithm according to the business logic,it finally meets the needs of the Grid Center.Based on multi-source data such as basic infrastructure information,this paper analyzes in detail the triggers that cause each type of incidents and the factors that cause the deterioration of various types of incidents.Based on the analysis of the correlation of triggers,this paper establishes multi-dimension fusion prediction model to predict the development of various incidents in the future.Based on the evolutionary factor correlation analysis,all kinds of incidents are monitored during their occurrence and evolution and are warned before they deteriorate.