Research on the construction and optimization of physical education teaching analysis platform based on Bi-LSTM model

With the extensive application of information technology in education, physical education teaching is gradually optimized and improved using data-driven methods. This paper focuses on constructing and optimizing the physical education teaching analysis platform by using the two-way long and short-term memory network technology. The study collected multi-dimensional physical education data from >500 students, and conducted in-depth analysis through the Bi-LSTM model, aiming to improve the accuracy of teaching evaluation. The results show that the platform has achieved significant progress in the automatic scoring system, and the scoring accuracy has increased to 92 %, a 20 % improvement compared with the traditional methods. The platform can also accurately predict the physical improvement of students, with an accuracy of 85 %, and real-time analysis of skills to master the progress and sports risks, providing strong support for personalized teaching. These results not only enhance the objectivity of physical education evaluation, but also provide teachers with rich data insight and help them to develop more scientific and personalized teaching strategies.