Multi-stage clustering with complementary structural analysis of 2-mode networks

• Main Authors: Todeva, Emanuela, Knoke, David, Keskinova, Donka
• Format: Conference Proceeding
• Language: English
• Subjects: Clustering algorithms; Clustering methods; Couplings; data categorization; Data visualization; Industries; industry classification; multidimensional scaling; multiple correspondence analysis; multistage cluster analysis; network analysis; pattern recognition; Portfolios
• ISSN: 2473-991X
• DOI: 10.1145/3341161.3344781

This paper offers a synthesis of a new analytical procedure based on the complementary use of a large number of methods and techniques for categorisation of objects, pattern recognition and for structural analysis. It represents an example of a functional clustering [1] and an extension to the 'posteriori methods' for clusterisation [2]. We call this approach Multi-Stage Clustering (MSC), as it applies cluster analysis methods at three distinctive stages. We present the MSC and demonstrate its application to a business dataset of 275 multinational corporations (MNCs), aiming to address the inherent weaknesses of existing industrial classification tools designed to capture diversification of firms. We evaluate the outcomes from the MSC using a combination of complementary methods for structural analysis and data visualisation, such as multi-dimensional scaling (MDS), network mapping (NM) and multiple correspondence analysis (MCA). The MSC is designed for the analysis of diversification patterns of MNCs, which can enable the measurement of group competitiveness and performance across these patterns, known as industry segments, or strategic industry groups (SIGs).