Research on Personalized Medical Intervention Strategy Generation System based on Group Relative Policy Optimization and Time-Series Data Fusion

With the timely formation of personalized intervention plans based on high-dimensional heterogeneous time series information has become an important challenge in the medical field today. As electronic medical records, wearables and other multi-source medical data are increasingly generated and diversified. In this work, we develop a system to generate personalized medical intervention strategies based on Group Relative Policy Optimization (GRPO) and Time-Series Data Fusion: First by incorporating relative policy constraints among the groups during policy gradient updates adaptive balance the individual gain and group gain distribution. To improve the robustness and interpretability of decision-making, the multi-layer neural network structure was employed to group code the patient characteristics. Secondly, for the rapid multi-modal fusion of multi-source heterogeneous time series, a multi-channel neural network combined with self-attention mechanism was employed for dynamic feature extraction, the key feature screening and aggregation were further achieved through the differentiable gating network. Finally, a collaborative search process was proposed to find the ideal candidate intervention strategy based on the combination of genetic algorithm and Monte Carlo tree search so that a global optimization of the candidate intervention strategies was achieved, which greatly enhanced the accuracy of the system as well as the system's personalization level. The experimental results show that model achieves significant improvement in aspects of accuracy, coverage and decision-making benefits of intervention effect compared with existing methods.