| At present,it is in the rising period of digital transformation in urban traffic.The large-scale application of big data analysis technologies makes the urban traffic management move to the big data process.As an essential member of big data analysis technologies,clustering algorithm has enabled the intelligent development of many fields.In order to improve the applicability of existing clustering algorithms in the field of urban traffic,this paper focuses on providing a clustering algorithm with a high degree of adaptability,which can show excellent performance under the background of big data analysis and forecast in urban traffic.The research results can obtain potential and effective traffic information,and provide service for the intelligent development and scientific decision-making of urban traffic.In this paper,a weighted K-means clustering algorithm based on time series is proposed.Based on the proposed algorithm,Holt prediction and anomaly warning algorithm based on weighted K-means clustering and MLR model based on weighted K-means clustering are further proposed.The three proposed algorithms are applied to the big data analysis and prediction of traffic in Beijing.The main contents and conclusions are as follows:(1)Since the importance of sampling points in different time segments is not exactly the same,the coefficient of variation is taken to assign weight for improving K-means clustering algorithm.The improved weighted K-means clustering algorithm is proposed to identify the representative traffic congestion patterns.Case studies are conducted based on the real-life traffic congestion index data in Beijing,covering six districts from January 1,2017 to December31,2017.The results illustrate that traffic congestion patterns are both temporal dependent and spatial dependent.(2)Considering the weighted K-means clustering algorithm can identify the characteristics of the historical patterns of time series data,Holt prediction and anomaly warning algorithm based on weighted K-means clustering is proposed.This algorithm is applied to pattern identification,and real-time prediction and anomaly warning of urban road velocity.A case study shows that the proposed prediction and anomaly warning method has good performance based on the real-life traffic velocity data of Beijing Expressway from April 3,2020 to May 20,2020.(3)Taking the historical pattern identified by weighted K-means clustering algorithm as an independent variable of multiple linear regression model,combined with the input of other independent variables,a multiple regression prediction model based on weighted K-means clustering is proposed.Based on the multi-source traffic data sets of Beijing West Railway Station from August1,2019 to October 15,2019 and from August 1,2020 to August 31,2020,this paper explores the travel characteristics and forecasts the traffic demand of traditional taxi and online car-hailing.The results show that the proposed prediction model has good prediction performance.The identification of urban traffic congestion patterns,the prediction and anomaly warning of traffic velocity in urban road network and the travel demand prediction of urban railway stations are interrelated with the urban traffic management system.This study proves the good application performance of weighted K-means clustering algorithm and its related algorithm.The case study results also have vital practical significance for formulating specific traffic optimization and control schemes,alleviating traffic congestion and improving traffic conditions. |
