Internet technology promotes the innovation of education mode, and OMO teaching mode deeply integrates online and offline teaching and lacks personalized learning path planning. The accumulation of largescale educational data provides a foundation for learner behavior analysis, and the construction of knowledge point difficulty model and learner state model through data analysis algorithms realizes personalized learning path planning for different learning styles, improving learning effect and satisfaction. In this study, cognitive network analysis (CNA), social network analysis (SNA) and content analysis (CA) are used to analyze and process educational data, focusing on learners’ online learning behaviors and practice test scores, and establishing a model for judging learners’ learning status. The experimental results show that the learning path planned based on the CNA method is more in line with the user’s learning style, and the similarity with the user 2 learning style reaches 0.90, which is higher than the SNA and CA methods. The data analysis showed that the number of library behaviors of students with good grades (about 19.73% of the total number of students) reached 52 times, which was significantly higher than that of students with average grades (46 times) and students with poor grades (40 times). In addition, seven key factors were extracted through principal component analysis, which could explain 69.945% of the overall variance, effectively reflecting the correlation between students’ behaviors and academic performance. The study proves that personalized learning path planning based on large-scale educational data analysis can effectively meet the needs of users with different learning styles, improve learning efficiency and user satisfaction, and provide effective methodological support for the practical application of OMO teaching mode.
