دانلود رایگان مقاله قدرت پیش بینی شبکه تطبیقی برای عملکرد بخش سهام

Abstract

As economic integration and business connections increase, companies actively interact with each other in the market in cooperative or competitive relationships. To understand the market network structure with company relationships and to investigate the impacts of market network structure on stock sector performance, we propose the construct of a company comparative network based on public media data and sector interaction metrics based on the company network. All the market network structure metrics are integrated into a vector autoregression model with stock sector return and risk. Several findings demonstrate the dynamic relationships that exist between sector interactions and sector performance. First, sector interaction metrics constructed based on company networks are significant leading indicators of sector performance. Interestingly, the interactions between sectors have greater predictive power than those within sectors. Second, compared with the company closeness network, the company comparative network, which labels the cooperative or competitive relationships between companies, is a better construct to understand and predict sector interactions and performance. Third, competitive company interactions between sectors impact sector performance in a slower manner than cooperative company interactions. The findings enrich financial studies regarding asset pricing by providing additional explanations of company/sector interactions and insights into company management using industry-level strategies.

6. Discussion and conclusions

This study aims to construct an effective company relationship network using big data and to investigate the dynamic relationships between sector interactions and stock sector performance. The results suggest that company networks constructed based on public news provide predictive indicators for sector performance and that inter-sector interaction has a stronger predictive power than intra-sector interaction. Moreover, in the network construction, comparative analysis provides a better method than closeness analysis. The negative interactions have a shorter reaction time than the positive interactions for return, and they have longer effects for both sector return and risk. These findings are also confirmed using the links as alternative metrics to reflect the interactions between sectors. Collectively, these findings provide important implications for research regarding market structure and stock sector performance.