THR-DMPOM model for reliable high-precision evaluation to enhance real-time performance management and optimize resource allocation

To address volatility and local optimality issues arising from multidimensional dynamic data in Human Resource Performance Evaluation (HRPE) within the tourism industry, this study proposes a Tourism Human Resource–Deep Momentum Performance Optimization Model (THR-DMPOM). The development of the model begins with the construction of a 24-item evaluation index system. This system was established through an extensive review of HRPE literature and interviewed with more than 300 managers and employees from four representative tourism enterprises. Following multiple rounds of expert screening, 24 key indicators were finalized. Methodologically, THR-DMPOM introduces recursive momentum accumulation, cross-layer weight smoothing, and adaptive gradient correction mechanisms. These components enable continuous dynamic weight updating and historical gradient propagation across multilayer networks, thereby improving stability in nonlinear and time-varying evaluation environments. Extensive experiments were conducted using datasets from four tourism enterprises. The results show that THR-DMPOM achieves high stability and accuracy in practical applications. The systemic deviation remained within ± 1.0, demonstrating strong robustness in handling complex performance data. In distribution consistency validation, the D-values of THR-DMPOM scores remained close to 1, with p-values consistently exceeding 0.7, indicating strong agreement between model-generated results and historical records. Comparative experiments under high-load and unstable network conditions further confirm the superiority of the proposed model. THR-DMPOM outperforms Fuzzy Analytic Hierarchy Process (FAHP) and the Fuzzy Technique for Order of Preference by Similarity to Ideal Solution (F-TOPSIS) in both accuracy and operational efficiency. Compared with traditional AHP, TOPSIS, and static machine learning methods, which struggle to adapt to dynamic performance fluctuations and nonlinear interdependencies among multiple criteria, THR-DMPOM demonstrates a key advantage in adaptive weight learning. The proposed mechanisms enable continuous weight evolution and historical information sharing across network layers, effectively mitigating local optima and weight oscillation issues in dynamic environments. Overall, the model provides a robust and high-precision evaluation framework for real-time performance management and resource optimization in tourism enterprises operating under dynamic conditions.