کنفرانس بین المللی محاسبات نرم

صفحه اصلی / ششمین کنفرانس بین المللی محاسبات نرم

Design of a Network-Oriented Framework for Detecting Key Stocks in the Tehran Stock Exchange via Time Series Analysis and Nonlinear Techniques

نویسندگان :

Fatemeh Nazemi Harandi¹ Zainabolhoda Heshmati Rafsanjani²

1- دانشگاه تهران 2- دانشگاه تهران

کلمات کلیدی :

Network Analysis،Transfer Entropy،Dynamic Time Warping،Tehran Stock Exchange،Centrality Measures

چکیده :

This study develops a network-based analytical framework for identifying key stocks and portfolio strategies in the Tehran Stock Exchange. The objective is to capture nonlinear and directional dependencies among stocks that traditional linear methods often overlook. Daily prices of 20 highly traded stocks from March 2016 to March 2024 were analyzed. Nonlinear relationships were quantified using Dynamic Time Warping (DTW) for dynamic similarity and Transfer Entropy (TE) for information flow, combined into a weighted directed network. Key stocks were identified through multiple centrality measures and the Louvain community detection algorithm. Two portfolio strategies—equal-weighted and centrality-weighted—were evaluated using a rolling backtest. Results show that the equal-weighted portfolio achieved higher risk-adjusted returns, while the network-based approach provided structural insights into hidden market dependencies.