Abstract: To address the limitations of existing portfolio research in capturing higher-order moment risks and nonlinear correlation structures, this paper develops a Higher-order Network Topology-Integrated Parametric Portfolio (HNT-PP) model from a complex network perspective. First, the GJRSK model is employed to measure conditional volatility, skewness, and kurtosis risks. Integrated with complex network and rolling window techniques, this approach dynamically depicts the multidimensional, nonlinear higher-order moment risk spillover network among commodities. Second, the eigenvector centrality is extracted from these networks and incorporated into a CRRA-based parametric portfolio framework to optimize asset allocation decisions. Empirical evidence from international commodities reveals a significant negative correlation between the network centrality of asset and its optimal allocation weight, implying that positioned assets receive substantially lower allocation weights than peripheral assets. Compared to competing models, the HNT-PP model delivers significant and robust improvements in terms of returns, risk, and risk-adjusted returns. Notably, the HNTVSK-PP model, which jointly integrates volatility-, skewness-, and kurtosis-based network topological characteristics, effectively captures dynamic changes in higher-order moment risk structures and adjusts allocation weights in a timely manner, thereby achieving superior portfolio performance. This study extends the application of complex network methods in portfolio management and provides valuable insights for dynamically optimizing commodity asset allocation in complex market environments.

Key words: commodity portfolio, higher-order complex network, network topology, parametric strategy