Time Series Traffic Analysis based on PCA and GMM Algorithm

In modern traffic systems, traffic flow analysis in different areas is the key to traffic flow prediction. Because the traffic flow data is essentially time series data, which has the characteristics of large dimension and long duration, it is impossible to extract its distribution characteristics directly. The traditional methods based on linear regression, autoregression and gray prediction have low accuracy, while the methods based on deep learning have high computational complexity. PCA algorithm can effectively separate high-dimensional mixed data into low-dimensional separable data. Compared with other unsupervised clustering algorithms, Gaussian mixture model can better estimate the distribution of original data. Through the simulation and comparison of standard clustering algorithms, it is found that PCA and GMM algorithms have higher accuracy for traffic flow model estimation and are closer to the distribution characteristics of the original data.