Abstract:

A prediction algorithm for multi-data stream based on gene expression programming(GEP) was proposed for compensating the shortage that the traditional linear regression method could only adapt to a simple prediction model and predict AVG in the future window. A method by using sliding windows to partition data stream was given in the multi-data stream. The main conception of Multi-Streams was defined, and the map relation in it was revealed. An algorithm was given to pretreatment multi-data stream according the map relation and the sliding windows above. An adaptive forecasting model was put forward based on DSMA-GEP in the multi-data stream. The experience with the real data showed that the method was efficient.

Key words: prediction model; gene expression programming; multi-data stream