Abstract: The identification of wool and cashmere is not only a difficult point in the fields of fiber detection but also the research mainly based on traditional statistical methods. So the classification research of wool and cashmere was made from the perspective of data mining here for the first time. By using multiple indicators on a single fiber as the characteristic attributes for the classification research, the fiber classification was explored from a new perspective. And the classification research was done between sheep wool and cashmere by using four main algorithms among the classical decision tree. And the relative mathematical modeling and evaluation was also completed. Experimental results show that the average relative error of built models is less than 6%. Through comparison, the C5.0 algorithm is more accurate and stable than other algorithms, and it can be used for practical classification and detection of wool and cashmere.

Key words: cashmere, wool, inspection, data mining, decision tree algorithm