Network Analysis of Technology Stocks using Market Correlation

In this paper, we propose to use network approaches to analyze correlation between stocks. Our essential goal is to directly answer four questions: (1) how stocks in certain industry sector are correlated to each other' (2) what are the characteristics of stock networks with respect to their market behavioral correlations, and (3) do stocks in an industry sector form meaningful groups, based on on their market behaviors based correlations, and (4) how robust a correlation based network analysis approach can be used to understand stocks as a graph. In order to provide clear answers to address the above questions, we used market correlation methods to generate stock graphs. Two community detection methods, Louvain Modularity and Walk Trap, were used to study the structure of the graphs. To further test the robustness of our model, we created another graph using different correlation threshold. In the experiment we detected twelve communities using Louvain Modularity method and they consisted of stocks from different industries. Even the smallest cluster, which included only 2-3 stocks contained stocks from different industries.