In K-12 intelligent learning environments, how teachers provide adaptive support affects student performance. This study employs a chain mediation model and learning analytics to explore how perceived need support affects learning outcomes in human-computer collaborative teaching. Results show that need support is significantly related to exercise engagement and that structure and involvement impact achievement through self-regulated learning and exercise engagement. Behavioral sequence analysis further reveals that structure and involvement shape students’ behaviors and strategy during exercise on platforms. This study offers theoretical insights into teacher-computer collaboration and practical guidance for optimizing instructional design and learning support.