RNA binding proteins play an important role in the process of post-transcription, identifying special RNA binding domains and interacting with RNA. Although many calculation methods have been proposed, most of them have the problems of insufficient features and unbalanced samples. This paper proposes a Stacking classification model composed of 4 base classifiers and 1 meta classifier. While enriching features, it extracts as much information as possible from different feature expression methods. We use the dipeptide distribution matrix to supplement the missing dipeptide position information in the amino acid composition. The sliding window method is used to balance the positive and negative samples, and the sequence length distribution is more reasonable. The results show that the Stacking classification model has a certain improvement in the accuracy of RNA-binding protein sequence prediction. At the same time, the position information contained in the dipeptide distribution matrix shows more excellent performance than amino acid composition information.