Multimedia-assisted data mining algorithms are used in this paper to extract information from soccer training teaching data and develop models for the sampling and storing soccer training data. Filtering is used to process the soccer training attribute data stream, and global position optimization is performed using the mutual information analysis method to integrate the teaching system resources. The soccer training teaching model and teaching program were established to finish the teaching system design. The results show that after one semester of teaching, the final assessment, compared to the average score of the entrance assessment, the training performance increased by about 42%, thus proving the effectiveness of the soccer training teaching system based on multimedia-assisted teaching.