Abstract: A Web data mining algorithm based on semi structure feature segmentation is proposed. The information stream signal model of Web hot date is constructed and the characteristic erwelope decomposition of Web hot information stream is finished, in order to improve the purity of data mining and the anti-interference performance by feedforward filter modulation data interference filter, using semi structural feature segmentation for web hot number according to feature extraction. The data mining algorithm is realized. Simulation results show that the new algorithm can improve the detection capability of characteristics of Web data, data mining has little sidelobe interference, mining precision is high, performance is better than traditional algorithm.