Optimization of Cross-regional Knowledge Sharing Mechanism for Community Infant Care Guidance Based on Deep Transfer Learning and BiLSTM

Abstract

The inefficiency of cross-regional knowledge sharing mechanism leads to the inability to effectively circulate and reuse valuable infant care guidance resources among communities, which restricts the improvement of infant care guidance services. This study aims to build an optimized cross-regional knowledge sharing mechanism and use deep transfer learning to transfer the deep patterns and knowledge contained in community infant care guidance data in knowledge-rich areas to resource-scarce areas. Secondly, a BiLSTM model is constructed to process the sequence dependency and contextual semantic information of infant care guidance text, effectively capturing the temporal features and key patterns in guidance suggestions. Then, a specific feature alignment strategy and domain adaptation module are designed to minimize the impact of the difference in data distribution between the source domain and the target domain on the transfer effect. Finally, the attention mechanism is applied to optimize feature selection and improve the model's recognition of core guidance knowledge points. In the cross-regional knowledge sharing task, this mechanism improves the accuracy of generating infant care guidance suggestions in the target community to 92.7% and the knowledge transfer efficiency to 35.5%, and the resource utilization rate of the target domain community reaches 88.8%. The optimization mechanism based on deep transfer learning and BiLSTM can effectively break through regional barriers and provide strong technical support for the level of infant care guidance services in underdeveloped areas.