Limited Tolerance Relation-Based Decision Tree Algorithm

In this research, we study how to generate a decision tree from dataset with unknown values, and proposed a decision tree learning algorithm (LTR-C4.5). The algorithm based on limited tolerance relation and C4.5. Algorithm LTR-C4.5 is composed by two function modules: filling the unknown values and generating a decision tree. The algorithm recursive calls the two function modules when handling incomplete training samples. The outstanding feature of LTR-C4.5 is that it doesn´t demand to fill all unknown values before generating a decision tree. Some experiments are used to simulation the algorithm and compared it to other methods.