To provide a method and device for preparing an effective and valid decision tree for executing valid classification.

In this method and device for generating a decision tree by using linear judgment analysis, and for executing the decision tree for data classification (categorization), it is desired that the data are constituted in the format of multi-dimensional objects, for example, the format of data records including characteristic valuables and class variables in a decision tree generation mode and data records including only the characteristics variables in a decision tree scanning mode. For example, a more effective classification system with monitor is prepared by this method. Generally, the recursive division of the decision tree is included in this method so that the maximum amounts of separation can be achieved among the class values of training data. This is executed by finding the effective combination of the variables in order to prepare the decision tree by recursively dividing the training data. This decision tree is used for classifying input test data afterwards.