AI-Powered Stock Forecasting: A Graph-Based Approach for NEPSE
DOI:
https://doi.org/10.3126/jonc.v1i1-2.89049Keywords:
Graph Neural Networks (GNNs), financial forecasting, Graph Convolutional Network (GCN), Graph Attention Network (GAT), stock market prediction, Nepal Stock Exchange (NEPSE)Abstract
The stock market is a cornerstone of the financial ecosystem, yet forecasting price movements remains a formidable challenge due to the dynamic and interconnected nature of influencing factors. While conventional prediction models often fail to adequately represent these complex relationships, Graph Neural Networks (GNNs) have emerged as a promising alternative, offering superior accuracy by modeling financial data as interconnected graphs.
In this study, we introduce a visibility-based graph transformation technique to convert stock market features into a structured network, capturing long-memory dependencies. We then apply Graph Convolutional Networks (GCN) and Graph Attention Networks (GAT) to analyze trends and predict market behavior. Our experiments reveal that GCN outperforms GAT in modeling financial graph structures, demonstrating its robustness in deciphering intricate market relationships.
These results underscore the potential of GNN-driven approaches in stock market forecasting, providing actionable insights for investors and advancing predictive analytics in the Nepalese stock market (NEPSE).
Downloads
