A Study on the Integration and Practice of Integrated Transfer Learning Strategies in Japanese Language Teaching and Learning

Addressing the prevalent issue of inefficiency in Japanese language learning, this study focuses on innovating classroom teaching methodologies for the Japanese language. Initially, the research employs a stochastic gradient descent algorithm to enhance the performance of a transfer learning algorithm, aiming to accurately identify student behaviors within the Japanese language classroom. Subsequently, these behaviors are quantified into integration weights using a predefined integration strategy. Behaviors demonstrating significant similarities are consolidated and analyzed to yield the results of the student behavior identification process. Furthermore, this behavior recognition approach is seamlessly incorporated into Japanese language pedagogy to develop a refined model for teaching, thereby potentially increasing the efficacy of language acquisition. Using this paper’s algorithm to identify students’ classroom behaviors and correlate them with their grades, it is found that when the number of positive behaviors in the classroom is higher, the student’s academic performance is also higher (r=0.9986, p=0.001<0.05). The average percentage of students’ attentive listening behaviors could reach 52.24% after the intervention, as determined by behavioral recognition results in the Japanese language teaching classroom. The behavior recognition algorithm in this paper provides an effective method for analyzing and intervening in students’ behavior in the classroom, and the teaching model constructed based on the algorithm offers a reference way to improve the effectiveness of Japanese language teaching.