The Research On KPI Association Relationship Discovery Method Based On Large Traffic Data

With the rapid progress of urbanization and rapid economic growth in China,large numbers of people and traffic flow are emerging in cities,resulting in large-scale,real-time and high-density data.These data contain rich city information and great economic value.The analysis of Key Performance Indicators(KPI)in city data helps to mine valuable information from data and provide a reliable basis for urban management and planning.Based on taxi GPS data,bus GPS data,bus IC card data,urban road data,points of interest data,and meteorological data,and the analysis of the relationship between traffic data,we study the two key performance indicators of transportation carbon emissions and bus trips.In this paper,we propose a three-layer perceptron neural network(3-layerPNN)and a deep neural network(SDAE-4)to predict the transportation carbon emission and bus trips throughout the city during the future period of time,respectively.Moreover,the representative features are extracted from the spatial-temporal urban datasets,such as taxi GPS data,bus GPS data,bus IC card data,urban road data,points of interest data,and meteorological data.Based on the future information of people's bus trips and transportation carbon emission in the whole city,we use a greedy mechanism to recommend the optimal bus routes for electric buses.Finally,we evaluate our approach through extensive experiments on the real-world data sources in Zhuhai,China.The results show that the proposed 3-layerPNN has advantages over the well-known three machine learning methods(Gaussian Naive Bayes,Linear Regression,Logistic Regression)and two deep learning methods(Stacked Denoising Autoencoder,Deep Belief Networks).Moreover,our proposed SDAE-4 deep neural network is consistently superior to five baselines,that is logistic regression,SVM with RBF kernel,restricted Boltzmann machines,artificial neural network,and deep belief network.