Deteksi Komunitas Pasar Saham IHSG dengan Metode Hybrid Jaringan Kompleks dan Algoritma Leiden

Abstract

The stock market is a complex system, with relationships between stocks that influence each other and form a dynamic network. In Indonesia, the Jakarta Composite Index (JCI) reflects the movement of the stock market as a whole. This study aims to detect the community structure of stocks in the JCI by sector using a hybrid approach that combines Random Matrix Theory (RMT), Complex Network (CN), and Leiden algorithm. The data used is the daily closing price of stocks in the JCI during the period January 2014 to January 2024. The methods applied include the formation of a correlation matrix between stocks, noise filtering using RMT, and community analysis using the Leiden algorithm. A multi-threshold correlation approach (0.7; 0.8; and 0.9) was used to evaluate the strength of the relationship between sectors. The results show that the combination of RMT, CN, and Leiden algorithm is effective in identifying stock communities with significant relationships. A higher correlation threshold results in a more stable community with a maximum modularity value of 0.72 at a threshold of 0.9. This approach makes an important contribution in understanding cross-sector interactions in the JCI stock market.