Research And Application On The Dynamic Fuzzy Decision Tree Learning

Decision tree learning is a extensive used learning method. This article caught the newly faced dynamic fuzzy problem of decision tree. Firstly, we used dynamic fuzzy lattice to describe question and build the decision tree on these dynamic fuzzy information. We called it dynamic fuzzy decision tree (DFDT).We bring forward the newly attribute chosen algorithm based on the domain of condition attributes. Secondly, we discreted the domain and the grade of membership base on the entropy of information. We also chose the inside junction point on it. We gave a method deal with those examples which lose some attributes. We designed a new system to prune DFDT by itself when it's working. After all, we designed a system of traffic monitor. It could be used to monitor the traffic of roads in cities. We researched the application of DFDT base on this system. All in all, features of the paper represent as follow:(1)We researched how to use the examples which lose some attributes. When DFDT is built, these special examples can be used rather than be thrown away. We also make DFDT can deal with input examples which lose some attributes.(2)We designed the system of pruning a DFDT by itself. This system makes a DFDT work better, if this DFDT work for a long time. So two DFDTs will become different trees gradually. Then they'll satisfy different people.(3)We researched the application of DFDT base on the system of traffic monitor. We also simulated this system with the computer program.Although we got some summings-up, the work what we do is basic research. There are some questions that need be deeply researched. For example, how to build a DFDT quickly, and how to make the structure of new DFDT better when it's pruning itself.