规则简化与模糊决策树剪枝的比较

决策树归纳学习算法是机器学习领域中解决分类问题的最有效工具之一。由于决策树算法自身的缺陷了，因此需要进行相应的简化来提高预测精度。模糊决策树算法是对决策树算法的一种改进，它更加接近人的思维方式。文章通过实验分析了模糊决策树、规则简化与模糊规则简化；模糊决策树与模糊预剪枝算法的异同，对决策树的大小、算法的训练准确率与测试准确率进行比较，分析了模糊决策树的性能，为改进该算法提供了一些有益的线索。

A Comparative Analysis of Rule Simplification and Pruning Fuzzy Decision Trees

Decision tree induction learns the implied rules from the training set, and then uses the learned rules to predict for unseen instances. However, the crisp decision trees often suffer from overfitting the training set in real-world induction tasks. So the pruning decision tree methods are necessary in the process of building crisp decision tree to improve performance. Fuzzy decision tree induction is an extension of crisp decision tree induction and is more close to the way of human thinking. In this paper, a comparative study is made among fuzzy decision tree algorithm, the simplified rules, and fuzzy simplified rules, fuzzy decision tree and fuzzy pre-pruning methods, with the aim of understanding their theoretical foundations, their performance and the strengths and weaknesses of their formulation. The empirical results show that fuzzy decision tree is superior to crisp simplified rules. The fuzzy pre-pruning decision tree can build a good tree even without simplified rules method.