Abstract: This paper introduces the disadvantages of two traditional data access modes for data mining algorithms，and then presents a new Cache-based data access mode.The mode supports three buffer patterns，which are single column pattern，multi-column pattern，mixed-columns pattern.These patterns are suitable for most of the data mining algorithms.Design and implementation of this Cache-based data-mining-specific data access component is presented later.It takes advantages of two traditional modes，ensures the high performance and supports the great capacity of data access with limited system resource.

Key words: Cache, data access, data mining, component

摘要： 介绍了数据挖掘算法的两种传统数据访问方式及其缺点，提出了新的数据访问方式——基于Cache的数据挖掘算法的数据访问方法，该方法提供了三种模式缓存数据：单列模式、多列模式、混合模式，以适用多种数据挖掘算法的需要。设计实现了这种数据挖掘专用数据访问组件，该组件兼顾了传统访问方式的优点，实验证明在占用有限系统资源的情况下，保证了高效的数据访问效率并支持对海量数据的访问。

关键词: Cache, 数据访问, 数据挖掘, 组件