Studies on Design and Development of Web Data Warehouse

随着万维网的迅速发展，有必要开发一种高效的检索和处理网络内容的方法。服务于此目的的两个众所周知的方法是web缓存和信息过滤。前者增强了检索数据的可重用性，而后者限制了对有价值数据的检索。在本研究中，将这两种方法进行了修改，然后将其统一为一个系统，并命名为“Web Warehouse”。在改进web缓存算法的基础上，提出了一种基于分类规则和分级内容管理模型的虚拟多层内容缓存模型。仿真实验证明，该LRU-SP模型的实现比常规系统具有更高的整体效率。同时，提出了一种利用超文本结构和基于关键词的web内容语义的方法，称为“基于内容语义的智能缓存方法”。该方法通过LRU-SP+实现。该研究还包括将web缓存数据修改为数据库，分析给定数据的特征，以及web数据仓库的动态分层结构。该结果在一次会议上发表，该会议限制了真正具有创造性的结果（在会议上的28份报告中，只有1份来自亚洲），并因其独特性而受到广泛关注。利用web内容的局部性特性来改进信息过滤。与地理信息系统（GIS）、自然语言处理和超链接分析的统一使得将网络内容限制为本地内容成为可能。该方法提高了局部信息搜索的效率。虽然Web最初是作为一个全球信息系统而构建的，但我们的系统也将其与本地信息搜索结合起来。这种方法在世界范围内受到了关注，韩国也正在开发同样的方法。

Due to the rapid expansion of the World Wide Web, it is necessary to develop efficient way of retrieving and processing web contents. Two well-known methods serving for this purpose are web caching and information filtering. The former enhances reusability of retrieved data, while the latter limits retrieval of data to worthwhile ones. In this research, two methods were modified and then united into a single system, under the name of "Web Warehouse".In improving web cache algorithms, a new model named Virtual Multi-Layered Contents Cache was developed, using classification rules and hierarchical contents management model. The implementation of this model, LRU-SP, was proved to have higher overall efficiency compared to conventional systems in simulation experiment. At the same time, a method using hypertext structure and keyword-based semantics of web contents, named "Intelligent Caching Method using Contents Semantics". This method was implemented as LRU-SP+. The research also contained modification of web cache data as databases, analysis on characteristics of the given data, and dynamic hierarchical structure of web data warehouse, The result was presented in a conference that limits presentations to truly inventive results, (out of 28 presentations in the conference, there was only 1 from Asia,) and has received wide attention for its uniqueness. For improvement on information filtering, locality properties of web contents were used. Unification with Geographic Information Systems (GIS), natural language processing, and hyperlink analysis has enabled restriction of web contents to local ones. This method has improved efficiency in local information search. Although the Web is originally constructed as a global information system, with our system it is also united for a local information search. This method has also gained worldwide attention, and systems with same method are now under development in Korea.