A Study of Driving Conditions Based on Principal Component Analysis and Optimization Clustering Algorithm

As the fuzzy C-means clustering algorithm is easy to fall into local optimum and the traditional principal component analysis (PCA) does not fully reflect replacement of multiple indexes with a small number of composite indicators, an improved principal component analysis and a genetic simulated annealing algorithm are proposed to optimize the fuzzy C-means clustering algorithm (GSA-FCM), so as to build the driving condition. First of all, the improved PCA is used to deal with the characteristic parameter matrix. Then, GSA-FCM clustering algorithm is used to cluster the kinematics fragments. Finally, the appropriate fragments are selected to synthesize the final working pattern. Moreover, the effectiveness of the characteristic parameters in the synthesis conditions of GSA-FCM clustering and traditional K-means clustering algorithm and the actual conditions were verified and compared with that of the standard test conditions of NEDC. The experimental results show that the average relative error of characteristic parameters between the condition synthesized with GSA-FCM clustering algorithm and the actual condition is 6.46%, which indicates that the clustering effect of GSA-FCM clustering algorithm is obvious and the error is small, and the synthesized driving condition can represent the actual driving condition of the city.