Accession Number:

AD1046410

Title:

Megacity analysis: a clustering approach to classification

Descriptive Note:

Technical Report

Corporate Author:

Naval Postgraduate School Monterey United States

Personal Author(s):

• Housley,Jimmy L.

Report Date:

2017-06-01

Pagination or Media Count:

153.0

Abstract:

Megacities are characterized by large populations at least 10 million and interdependent infrastructure, demographic, economic, and government networks the four pillars. To be successful in future operations, the military must expand its understanding of megacities and their networks. In particular the Joint Warfare Analysis Center JWAC is interested in these megacity networks and their implications for potential urban operations. We develop a methodology to group like megacities into five clusters. With 33 variables describing the four pillars, we construct a data set using over 90 data sources for 41 large urban areas. This work greatly expands previous work in both the number of cities studied and the number of variables used. We also study clustering sensitivity to missing values by generating an ensemble of 5,000 clusterings based on randomly imputed missing values. We compare these to clustering without imputation, the ensemble consensus or average clustering, and clusterings from previous studies in addition to identifying which cities are sensitive to missing values. Our work not only informs JWAC of the similarities and differences between the 41 cities studied, it provides a method to identify for which cities, more data collection is warranted, and it provides a blueprint for future work in this area.

Descriptors:

• urban areas
• POPULATED PLACES
• unsupervised machine learning
• clustering
• network analysis (management)
• classification
• demography
• infrastructure
• cities

Subject Categories:

• Operations Research

Distribution Statement:

APPROVED FOR PUBLIC RELEASE