一种基于哈希链表的高效概念漂移连续属性处理算法

An Efficient Continuous-Valued Attribute Handling Algorithm for Mining Concept-Drifting Data Streams Based on the Extended Hash Table

本文重点研究了数据流挖掘中存在概念漂移情形的连续属性处理算法。数据流是一种增量、在线、实时的数据模型。VFDT是数据流挖掘中数据呈稳态分布情形下最成功的算法之一;CVFDT是有效解决数据流挖掘中概念漂移问题的算法之一。基于CVFDT,本文提出了有效地解决数据流挖掘中存在概念漂移情形的连续属性处理问题的扩展哈希表算法HashCVFDT。该算法在属性值插入、查找和删除时具有哈希表的快速性,而在选取每个连续属性的最优化划分节点时解决了哈希表不能有序输出的缺点。

This paper focuses on continuous-valued attribute handling for mining concept-drifting data streams. Data stream is an incremental,online and real-time model. VFDT is one of the most successful algorithms in data stream mining when data take on a state of stable distribution;CVFDT is one of the effective algorithms for resolving the problem of concept drifting in data stream mining. Based on CVFDT, the paper proposes an efficient continuous-valued attribute handling method named Hash CVFDT for mining concept-drifting data streams based on the extended hash table. The algorithm is as fast as the hash table in attribute inserting, seeking and deleting, and solves the flaws of the hash table which cannot output. Sequently when selecting the opthnally partitioned nodes of each continuous-valued attribute.