This paper takes constructivist learning theory as the theoretical guide and proposes an individual learning characteristic model that supports schema knowledge. A domain knowledge network construction method for learning path recommendation is designed, and the definitions of knowledge points and their relationships are systematically elaborated. Based on the two learning objective selection criteria of contribution degree and learning cost, two different learning objective selection algorithms are designed. With the learning objectives determined, an adaptive knowledge point learning sequence planning algorithm is proposed based on the type of learners’ learning styles. Taking the learning behavior data of 890 active learners on the Coursera platform as a sample, the empirical study of adaptive learning path recommendation is carried out. The optimal learning path recommendation is realized by synthesizing the results of tracking learners’ knowledge status in the knowledge network. The algorithm generates stable and optimal learning paths after 220 iterations in a linear algebraic knowledge network with the main interval of semantic distance values of [0.5,1.5]. The knowledge tracking results show that student 1 has a better mastery of A1, A2 and A3, partially learns A4, A5, A6 and A7, and basically fails to master A8 among the 10 learning records. When the learning gain, education, group, and technology weights of student 1 are 5, 2, 5, and 6, respectively, the corresponding learning paths are A1→A3→A4→A6→A8. The algorithm is able to automatically generate the optimal learning paths according to the available learning resources and learners’ knowledge levels. The algorithm can automatically generate learning paths according to the available learning resources and the knowledge level of the learner, and can adaptively adjust the learning paths in combination with the personalized needs of the learner.
