| 摘要 |
PROBLEM TO BE SOLVED: To provide a decision tree creating method using similarity index capable of reducing calculation amount in a program for creating a decision tree for the two-class sorting of real number data by solving the problem in a conventional method for creating a decision tree by using multi-dimensional data as they are that the calculation amount is remarkably increased according to the increase in dimensional number of data. SOLUTION: The similarity index between data is defined by an angle formed when the data are regarded as vectors, and how much each of the data is dependent on positive and negative classes is determined as the average value of similarity index between each of the data and the data group belonging to the positive or negative class. Since original data are referred to only once, the calculation amount can be reduced. COPYRIGHT: (C)2004,JPO
|
