Abstract:

Since FCM clustering was relatively sensitive to the initial clustering center, iteration was inclined to fall into local extremum and the global optimum was difficult to obtain, a modified FCM method was presented to overcome the above defects. In order to get closer to the global optimal clustering, the data of the principal components were classified by a SOM network, and the obtained weights were used as the initial clustering center of the FCM clustering. The modified FCM clustering method was used for establishing driving cycle in Hefei city. The theoretical analysis and its corresponding results indicate that this method possesses a sound precison for establishing driving conditions, which can reflect the realistic urban traffic conditions comprehensively.

Key words: fuzzy C means(FCM) clustering, self-organizing maps(SOM), principal component analysis, driving cycle

摘要：

针对FCM聚类法对初始聚类中心比较敏感、迭代容易陷入局部极值、难以取得最优聚类的问题，提出了一种改进的FCM方法,即利用SOM网络对主成分数据进行聚类,将得到的权值作为FCM聚类的初始聚类中心,从而使聚类结果更加接近最优聚类。将改进的FCM聚类方法应用于合肥市道路行驶工况的构建中,理论分析及试验结果表明，该方法有效地提高了聚类精度,构建的行驶工况与实际道路的交通状况吻合很好。

关键词: 模糊C均值聚类, 自组织映射, 主成分分析, 行驶工况