多策略通用数据采掘工具MSMiner

摘要

The design and implementation of MSMiner, a general multistrategy data mining tool, is proposed in this paper. MSMiner is founded on the basis of data warehouse, and integrated with several kinds of data mining algorithms, such as decision tree, association rule, statistical method, clustering, neural network and visualization. Also put forward is an object-oriented model to express and process metadata about data sources, algorithms and parameters, steps, tasks and users' information in data mining. The traits of MSMiner are the support of multiple data sources, i.e. data in database/warehouse, text and Web pages, the flexibility of the organization of data and mining strategies, and the powerful expandability of data mining algorithms and tasks.%介绍了一种多策略通用数据采掘工具MSMiner的设计与实现.MSMiner建立在数据仓库之上，采用面向对象的方法描述关于数据源、采掘算法、采掘步骤和用户的元数据.该系统集成决策树、关联规则、传统统计分析、聚类分析、神经网络和可视化等多种数据采掘算法，以任务模型的形式生成和执行数据采掘及决策支持任务.其特点是支持数据库、数据仓库、文本以及Web页面等形式数据源，可以动态地添加采掘算法，对数据和采掘策略的组织灵活有效，具有很好的可扩充性和通用性.