Enhanced Financial Portfolio Optimization with Risk Management using the GraphSAGE-PPO Model

Financial portfolio management is a longstanding yet increasingly complex problem in modern finance, characterized by high volatility, dynamic inter-asset dependencies, and the need for timely decision-making. Traditional methods often rely on static models and expert-driven heuristics, which struggle to adapt to changing market conditions and require significant human oversight. Moreover, many existing machine learning (ML) and deep learning (DL) approaches inadequately capture the relational structure among financial assets, leading to suboptimal portfolio allocation strategies. This study presents a fully automated portfolio optimization framework that works without manual input. It introduces a new model that combines GraphSAGE, a graph neural network for analyzing relationships between financial assets, with the Proximal Policy Optimization (PPO) algorithm, a reinforcement learning technique. Financial markets are modeled as graphs, and every asset is a node, and the relationships among them are indicative of price correlation. This structure enables the model to take into consideration the nature of individual assets and their relationship with each other. As compared to conventional techniques, the proposed strategy has a higher performance in returns as well as management of market fluctuations and has a well balanced portfolio.