Decision Tree Constructing System Based on Design Patterns
Maoqing Li · Journal of Xiamen University · 2005
Decision tree is an important way of classification in Data Mining.Decision tree has some basic algorithms,such as ID3 and C4.5.CART and PUBLIC algorithms are two important algorithms.Comparing these algorithms,we can know that there are some local changes affecting the construction of decision tree.This paper concludes a lot of“hot spots” in decision tree algorithm,such as data extracting,decision tree constructing,missing value handling,test attributes criterion selecting,decision tree pruning and decision tree describing,which include the local changes in algorithm fore mentioned and the design that should have good extensibility.Design patterns are experiences of software developments and designs.Applying design patterns to handle“hot spots”,the designs and codes of decision tree constructing system only revises little when these“hot spots”changed.Thus,this paper applies design patterns to design decision tree constructing system.Obviously,this solution has good extensibility and reusability.