This paper constructs a road network model after analyzing the structure of the English learning road network under the background of big data. After that, we analyze the characteristics of students’ learning behaviors by extracting them from their daily behavioral data and generalizing them using a large amount of data. The correlation between students’ campus behavior and academic performance is examined using the learning behavior road network model algorithm. Predicting students’ learning styles through the predictive assessment of their behavior and performance using a road network model. Through the clustering of students’ learning behavior portraits, we get the students’ portrait style, optimize the teaching implementation process in a targeted way, and realize the purpose of analyzing the behavior of English learners and optimizing the teaching strategy under the big data environment. The percentage of students who borrowed books 15-20 times (19%), 20-25 times (11%) and >25 times (3%) decreased significantly in the last two months. Most of the students visited the library 60-80 times, and most of them spent 150-200 hours studying in the library. There is a strong correlation between borrowing specialized books and students’ study hours and exam results. In addition, according to the characteristics of students’ study behaviors, this paper classifies them into five categories: “active (38.46%), self-disciplined (21.87%), risky (3.37%), routine (27.31%) and supervised (8.99%)”. Since different student portrait styles have their differences in behavioral characteristics, different teaching strategies should be adopted for students with different learning styles.