基于PSO改进决策树算法的研究

决策树方法是发现概念描述空间的一种特别有效的方法，是实例学习中具有代表性的学习方法，专门用于处理大量对象．如何快速建立简单可靠的决策树是一个重要的问题．文章引入PSO算法，并针对标准PSO算法易限于局部极小点的局限性，在保持了PSO算法结构简单可行特点的同时，利用惩罚函数方法，引入叉乘控制项，帮助算法摆脱局部极小点的束缚，提高了优化速度．将改进的PSO引入到决策树建树方法中，并与传统的决策树方法及使用遗传算法改进后的树进行比较，验证了其优越性．

Research on Decision Tree Method Based on Improved PSO

The decision tree method is the effective method of detecting for concept describing space and the representative learning way in exampling learning of which specially dispose mass object. Then how to establish the decision tree of simple and credible becomes the important problem. The paper introduces the Particle Swarm Optimize Algorithms, by adopting the method of punish function and the forking product controlling item, to conquer the shortcoming of the algorithm for getting into the scope around of local particle point. Building up decision tree by improved PSO, the paper gives the example to validate that the improved algorithm is better than the original decision tree method and by improved by GA.