Research On The Multidimensional Time Series Mixed Data Mining Forecasting Model Of Urban Traffic Flow And Its Application

Urban traffic congestion is the prevalent problems of the modern urban development,traffic forecast is the technology infrastructure of traffic management use the traffic guidance to relieve traffic congestion.At present,short-term forecasting of traffic flow theory research and technology is still in the development stage,yet the lack of authority of the theory.Build an accurate,real-time,fast traffic flow forecasting method is still an important issue in the current intelligent road traffic study domain.By studying various types of classical algorithms,combining the characteristics of traffic flow data,and considering the nearly road traffic multi-time series,this paper presents a new short-term traffic flow hybrid algorithms mode through multi-dimensional time-series mixed data mining.The main idea of the hybrid algorithm model is studying the nature of the timing characteristics of traffic data characterized,combining the global multidimensional time series and the classic method of machine learning.Firstly,time series data of the traffic flow is similaritily clustered,changcing multi-dimensional time-series data into two-dimensional information system to achieve an effective reduction dimension.Secondly,using the different classical machine learning program for data mining to get a different prediction information from the two-dimensional information table data of the first part.By comparison with the test data,determining the best prediction knowledge and forecasting programs.The specific embodiments of this model including: cleaning,filling,transposition,clustering,classification,machine learning,classifier structure,and the final prediction.The paper focuses on traffic data models: data present local traffic flow time series data model and the global traffic flow time series data model.Traffic flow data cleansing are discussed and practiced.After the road's traffic data are collected,multiple sets of comparative experiments show that the algorithm can select the optimal machine learning methods.It is proved that the algorithm is superior to the traditional single-path time-series data analysis method.