SPRINT算法的改进

An Improvement of SPRINT Algorithm

连续值属性的处理方法是影响决策树构造的重要因素。该文介绍了SPRINT算法,针对其最佳分割点的计算量大的缺点,提出了一种改进的快速寻找方法。该方法采用区间评估、筛选和局部逐一搜索等策略,大幅度地缩小了搜索空间。

How to deal with co ntinuous attributes is an important factor in decision tree construction.We f irstly introduce SPRINT algorithm,then present an exact method that finds the best split point quickly.The method uses a series of strategies such as interv als evaluation,filtration,and searching each point locally,narrow the search ing-space largely.