With the aim of providing personalized learning paths and targeted recommended learning resources for each student, this study is based on the web and designed for online English teaching platforms in universities. The hardware part of the platform is connected by multiple components through high-speed buses and network interfaces. The software part has designed a database that covers multiple roles such as students, teachers, and administrators, and has established a database permission management mechanism based on entities, attributes, and relationships. Based on the constructed database, the K-means clustering algorithm is used to standardize the Z-score of the original online English teaching data, and through clustering processing, student learning features are mined to achieve targeted recommendations. During the experiment, when 800 students were online at the same time, the page loading time of the platform designed in this paper was only 1.8 seconds, the memory utilization rate was 55%, and the CPU utilization rate was 72%. After applying the platform, the teaching effect satisfaction index was highly recognized, and the students’ English learning performance improved steadily. In the comparison of interface response speed, the dynamic frame rate of the platform increased from 2.0fps to nearly 3.0fps, which was a significant advantage, indicating that the platform has high application value.
