Matrix organizations are widely adopted for enhancd management efficiency through dual management structures. However, its complexity, intertwined hierarchical relationships, and diverse personnel mobility pose significant challenges for human resource scheduling. To effectively enhance the efficiency of human resources in matrix organizations, this study investigates multi-objective optimization models and algorithms for human resource scheduling in matrix organizations, fully considering the connections and conflicts between over/under-staffing rates and promotion rates across different units. Firstly, the matrix organizational structure and personnel mobility are systematically represented by a flow network, and a multi-objective optimization model is constructed for human resource flow networks in matrix organizations. Then, addressing the issues of insufficient convergence efficiency and poor uniformity of Pareto frontier distribution in traditional multi-objective optimization algorithms when processing complex problems, this research proposes an adaptive operator selection NSGA-II algorithm, which dynamically selects operators by the deep Q-network. Finally, based on the realistic characteristics of matrix organizations, test cases of varying scales are generated. By comparing this algorithm with other classical multi-objective optimization algorithms, its computational performance within the proposed model is validated. This research provides a scientific basis for decision-making in human resource scheduling problems in matrix organizations.