In view of the fact that the existing decision tree algorithms have not considered the time efficiency of rules produc-tion, this paper proposed a fast inverse decision tree algorithm for classification to improve the performance of decision tree al-gorithm. The algorithm began from the target of classification and aimed to produce accurate rules as quickly as possible. Estab-lished a new standard for classification, which mainly considered the speed of examples deflection due to different attributes. Designed a generator to produce random rules and data sets in the experiment. The experimental results show that the fast in-verse decision tree algorithm FIDT deduces the same rules faster than ID3. This induction and reasoning algorithm is especially feasible in applications with demand of time efficiency such as industrial production and system dispatching.%针对当前决策树算法没考虑规则生成时效的情况,提出了一种从目标函数出发,快速生成规则的逆向决策树算法,以提高决策树算法实时生成规则的能力.该算法采用了一种新的分类性能度量标准,该标准主要考虑不同属性对应的样本分布偏置的快慢.实验部分设计了一个随机规则和样本的生成器.实验结果表明逆向决策树算法拥有比ID3算法更好的时间性能和相当的规则生成能力.该归纳推理算法尤其适用于工业生产、系统调度等对系统实时性要求较高的社会经济与信息化领域.
展开▼
