Visualizing the clustering of financial networks and profitability of stocks

Abstract

We propose an approach to visualize the clustering of financial networks and a long-term profitability of stocks using financial time series data, by combining several methods of quantitative analysis. For demonstration purposes, this method is applied to investigate the network of Dow Jones Industrial Average (DJIA). Based on the time series data of stock prices during 31 July 2007 to 18 July 2011, our classification method clusters the DJIA components into five groups according to their profitability and property. By comparing the time correlation in the adjusted close price of stocks within the same group, we show that our clustering method results in a better classification of DJIA components than the methods of industry clustering and hierarchical clustering. With this integrated method, we have constructed a two-dimensional map of the DJIA network for visualization, and have related the first and second coordinates of DJIA components in the map to, respectively, their long-term profitability and property. Our analyses show very strong correlations for the sectors of Energy, Basic Materials, Technology, Capital Goods and Consumer/Non-Cyclical, significant correlations for the sectors of Services and Financial and a poor correlation for the Healthcare sector.

Visualizing the world's scientific publications

“…For large networks, MSC enables the users to view the network structure at various characteristic resolutions. Third, compared to the results of other clustering methods, the MSC clustering results have higher intragroup similarity and lower intergroup similarity in the social science network (Chang & Chen, 2011), the stock network (Chen & Chang, 2015), and the protein similarity networks (Hu et al, 2015). Finally, the MSC clustering algorithm is simple and easy to implement.…”

“…MSC is a simple and efficient algorithm and does not require predetermined inputs on the number or size of the clusters. The validity of MSC results has been tested for several complex networks (Chang & Chen, 2011;Chen & Chang, 2015;Hu, Mai, & Chen, 2015).…”