[Purpose/significance] Cultivating new quality productive forces presents new demands for accurately assessing the development trends of sci-tech innovation. It is an important tool for sci-tech development planning to pool
massive sci-tech information resources, deeply track the weak phenomena and interactions characterized in the early
stage of sci-tech innovation, and foresee the trends of emerging technologies evolving to frontier or core technologies.[Method/process] This study conducted a literature review on technology foresight analysis, explored key challenges in emerging technology foresight from the perspectives of evolutionary mechanisms and analytical methodologies, and constructed a graph learning-based emerging technology foresight framework from the weak signal perspective by integrating rules for judging the evolutionary stages of emerging technologies with representation learning from dynamic multiplex knowledge networks. [Result/conclusion] This framework employs a full-cycle analysis and revelation of evolutionary features and representational factors to deeply dissect the evolution mechanisms of emerging technologies.It proposes detailed steps for tracking and modeling the diverse evolutionary features of emerging technologies using dynamic multiplex knowledge networks, holding significant theoretical and practical implications for the field of sci-tech policy and management.