Feature Extraction Algorithm and Optimization for Mass TV Audience

User viewing feature can be extracted from TV user's channel viewing data for improving the targeted TV advertising. In this paper, we propose a basic viewing feature extraction algorithm in small sample environment to prove the algorithm logic and check the analysis process quickly. However if the viewing behavior data come from mass TV audience, it requires higher speed feature extraction algorithm. Therefore, we have optimized our algorithm based on four methods including data reading mode, data type, data table structure and data structure. After optimizations, the data processing efficiency rise 67 times.