Decision tree is the most important classification algorithm in data mining. At present, there are many decision tree algorithms, ID3 algorithm is the core one. This paper first studies and analyses the ID3 algorithm, then discusses the complicacy of computing the Information Entropy of attribute, and put forward a new heuristic based on the sensitive of attribute contributing to the classification. Finally, this paper compares the two algorithms by experiments, the results show that SID3 can generate the correct decision tree and the process is more simple, more quickly.