The Wiener process method is used to construct the degradation model, and the offline parameters are iteratively optimized using the EM algorithm combined with the training engine data. Based on Bayesian method and test engine data, an evolutionary data stream clustering algorithm (I-APDenStream) based on the fusion of nearest neighbor propagation and density is proposed. This algorithm uses the traditional two-stage processing model, namely online and offline clustering, to find the differences between English and Chinese general pronouns, the similarity and difference between the mother tongue and the target language, and the generalization of the sentence pattern or usage of the mother tongue. Use negative transfer. It aims to fully understand and scientifically grasp the status and role of language transfer in second language acquisition.