Adaptive online portfolio selection incorporating systematic risk of the financial market

Portfolio selection is a fundamental challenge in finance that aims to determine the optimal strategy for allocating capital among various assets. In the context of real-time investment activities, where market information is continuously updated, online portfolio selection has gained significant attention for its ability to efficiently derive optimal investment strategies. However, much of the existing literature tends to focus on maximizing returns while neglecting associated risks. In this paper, we first employ distance correlation coefficients to screen a set of historically similar samples. From this foundation, we construct a systematic risk indicator using a complex network and utilize the Adaptive Online Moving Average (AOLMA) method to predict future returns. The objective function maximizes cumulative returns while incorporating both systematic risk and transaction costs, which we transform into a linear programming problem, supported by theoretical proof. We introduce the Adaptive Online Portfolio Strategy (AOLNET) based on stock correlation networks. Comparative experiments demonstrate that the final wealth achieved by the AOLNET model, which incorporates similar historical samples, surpasses that of the AOLNPM model, which does not consider the risk indicator. This finding confirms that leveraging similar historical samples enhances final cumulative returns and that accounting for systematic risk effectively mitigates investment risk. Furthermore, numerical experimental results indicate that the algorithm proposed in this paper outperforms other widely used online portfolio strategies, highlighting the AOLNET model’s effectiveness in balancing investment returns while controlling risk.