Page 124 - Contributed Paper Session (CPS) - Volume 1

CPS1196 Song X. et al.
because of their historical reliance on secondary industries. Group E can be
called emerging industrial centers.

Figure 5: Normalized weights of secondary and tertiary industries for group E cites.

4. Discussion and Conclusion
In this research, we use the time-series cluster analysis to reveal the distinct
patterns in Chinese cities' economic growth. In particular, dynamic time
warping algorithm is used to measure the time-series distance between two
economic growth paths.
Using the weights of secondary and tertiary industries as input variables,
the time-series clustering has grouped the 35 major Chinese cities into five
categories. According to the industrial composition and the growth trend of
each of the five categories, they are labeled as service centers,
deindustrializing cities, balanced-industrializing cities, traditional industrial
centers, and emerging industrial centers, respectively.
The city categorization as revealed by clustering algorithm shed additional
light on how Chinese cities have developed over the last decade. Seemingly
distinct cities are shown to share more economical similarities than researchers
have thought. Moreover, with the role each city plays become clear,
competing and compensating relationships among cities are easier to identify.
This research is a preliminary study on how the unsupervised machine
learning techniques can be applied on the field of development economics. It
can be extended in various interesting ways. First, more metrics can be
included in the cluster analysis, so finer and more complete pictures of
economic growth paths can be painted. Second, fuzzy clustering techniques
can be applied, since it is common that a city would play multiple roles in an
economy. Finally, the economic network among cities is crucial to their
development; including graph theory based metrics that capture the network
effects into the model would provide additional insight.

113 | I S I W S C 2 0 1 9