基于动态时间规整的面板数据聚类方法研究及应用

以往的面板数据聚类方法存在一些缺陷，有必要基于动态时间规整的思路进一步改进与完善面板数据聚类方法。利用国家级经济技术开发区数据开展的实证分析结果表明：新的方法既能够很好地反映面板数据的动态变化、又避免了已有的面板数据聚类方法中各种距离如何赋权的问题，聚类结果较为稳定且有很好的可视化效果。

Research and Application of Panel Data Clustering Method Based on Dynamic Time Warping

Previous method on panel data clustering has some defects, so it is necessary to further improve and perfect panel data clustering method based on Dynamic Time Warping. This method is used to analysis the data of China national economic and technological development zone. The result of the empirical analysis shows that the new method not only reflects the dynamic change of panel data very well, but also avoids the problem of weighting the various distances in existing panel data clustering method. The clustering result is relatively stable and has a good visual effect.