Analysis Of Urban Hot Spots Based On Artificial Intelligence

With the social progress and rapid economic development,the process of urbanization is gradually accelerating.At the same time,urbanization can promote social progress and economic development.The degree of urbanization is an important indicator for evaluating the comprehensive competitiveness of a city.Urbanization brings convenience to people’s daily life,but it also brings problems related to urban planning.The spatial distribution of urban hotspots can represent the degree of urbanization.Therefore,analyzing the travel behavior of residents and extracting the urban hotspot travel areas,we can understand the spatial and temporal concentration of residents,urban spatial distribution and urban public issues.The identification of urban hotspots can help transportation planning managers to guide the evacuation of crowds and avoid traffic congestion,which provides convenience for residents to travel.Research on identifying urban hotspots is playing an increasingly important role in the study of urban issues.Urban hotspots analysis refers to the analysis of residents’ travel behaviors and the spatial distribution of the city by extracting the relevant data of residents taxis travels,providing more reasonable travel plans for residents,and providing more effective decision-making analysis for reasonable urban planning program.This paper uses data from Didi Chuxing’s "Gaia Data Open Plan"(https://gaia.didichuxing.com).By preprocessing taxi data,we obtain relevant data on Chengdu residents’ travel,and use artificial intelligence related technologies on urban hotspots analysis,and study travel areas based on massive taxi trajectory data.The main work of this paper is as follows:(1)Data pre-processing.The driver’s trajectory data of the Didi Express platform in Chengdu,65 square kilometers in October 2016 were used in this paper.In order to avoid the interference of invalid data with the analysis results,abnormal data need to be excluded.The GEN-SVM(Genetic-Support Vector Machine)classification algorithm is proposed by this paper.First,the genetic algorithm is used to optimize the important parameters in the Support Vector Machine(SVM)algorithm,and the SVM after the optimized parameters is used to classify the data;Second,the classification is used to judge the pros and cons of the classification results at intervals;Third,it determines the number of iterations of the algorithm by comparing the pros and cons of the classification results until the best(or better)classification results are obtained,stopping the algorithm iteration and outputting the final classification results;finally,road matching algorithm is used to match the normal data to the corresponding road sections,which lays the foundation for the analysis of the hot spots of residents’ travel.(2)Extract the OD(Origin departure point,Destination)information of residents’ travel from the trajectory data.The data of October is divided into the data of National Day,weekdays and weekends,in order to reduce the analysis deviation of a certain day due to weather,environment and other special reasons.For the impact,the values of working day and weekend data are the average of the corresponding data of working days and weekends from October 8 to October 31.The paper analyzes the number of trips and the length of travel during the week,weekdays,and weekends of National Day,compares the differences in holiday and work sunrise schedules of Chengdu residents ’travel activities,and visualize the distribution of residents’ travel in time.(3)A SC-GBSCAN(Silhouette Coefficient-based spatial clustering of applications with noise)clustering algorithm based on silhouette coefficient is proposed,which improves the problem of a large number of adjustment parameters of traditional DBSCAN algorithm in traffic data application.The algorithm firstly meshes the pick-up point area,maps the position information of the trajectory data to the grid,calculates the number of position data in each grid,and calculates the distance between the position data points;then based on the grid The range of internal data determines the Minpts parameter value in DBSCAN,and the minimum radius in DBSCAN is determined based on the distance between points,so that the parameter value could be within two ranges;finally,the clustering effect is determined by the contour coefficient and an iterative process is generated until there is a better clustering result to output the clustering result.In the proposed method,the clustering effect is determined by analyzing the original sample information,which simplifies the manual parameter adjustment in the original DBSCAN algorithm,improves the clustering efficiency,and visualizes the clustering effect on the page,which could visually see the residents’ travel The distribution of hot spots in space.