With the continuous advancement of digital teaching, English reading teaching in colleges and universities has put forward higher requirements for reading process identification, strategy analysis and accurate feedback. Aiming at the intelligent analysis task of college English reading comprehension strategies, this paper constructs a knowledge graph assisted analysis model, establishes a heterogeneous semantic network around reading text, topic type, behavior log and strategy label, and fuses text semantic features and learning behavior features. Gated fusion, graph reasoning and deep learning coding are used to realize strategy recognition such as skiming, scanning, inference and generalization. At the same time, the feedback adaptation, portrait update and joint optimization mechanisms are designed to form a closed loop of “recognition-feedback-correction-re-identification”. The experiment is carried out based on 168 reading texts, 840 questions, 4120 strategy samples and 28640 behavior fragments. The results show that the accuracy of the model is 91.9%, the F1 value is 91.0%, the F1 value of the logical inference task is 90.2%, and the average response delay is 0.67 s. The results show that the knowledge graph driven multi-source fusion and graph reasoning method can effectively improve the accuracy, adaptability and teaching support value of college English reading comprehension strategy analysis.
