Preference cost sensitive decision tree construction method

The invention discloses a preference cost sensitive decision tree construction method. A user can set the preference category, the preference degree and the performance cost matrix according to own preference; a candidate attribute list of an input training sample set is traversed, and a split attribute selection factor of each candidate attribute is calculated; a candidate attribute with the largest split attribute selection factor is selected as an expansion node of a preference cost sensitive decision tree; one branch is generated under the node corresponding to each attribute value of the candidate attribute with the largest split attribute selection factor; a sample set of each branch comprises all training samples of which the attribute value equals the corresponding branch; and the above steps are executed in loop until all samples in one node are empty or belong to the same attribute. The invention is used for balancing subjective preference for the preference category of a decision maker in the decision tree construction process so as to solve the problem of cost sensitive decision in the preference condition.