Research And Application On Link-Based Data Warehouse And Online Analytical Processing

Proposed in the early 1990's, data warehouse and OLAP were applied in a lot of fields in the middle 1990's. And now they have been important technologies to support the business decision making. In order to make the subject-oriented analysis easier, in the traditional data warehouse, the data about 'links' or relationships among two subjects has been transformed into the data which is focused on only one single subject. And this makes 'links' extraction very difficult. As the datasets are getting more and more complicated, newly emerging multi-relational data mining, graph-based data mining, link mining, etc. have become objects of interest for the academic world and the need of extracting links from the data warehouse has become more and more intensive. In order to mine these richly structured, heterogeneous datasets, link-based OLAP technology is also needed.In order to make link analysis easier, in this paper we firstly propose a new way to store these 'links' and extend the traditional multi-dimension data model and give the interpretation of link-based OLAP. Then, we design the architecture for the link-based OLAP system and give the description of each module. At last, we apply this technology to the telecom industry and implement a prototype system. By using the technique of the visualization of complex network to present the result in this prototype system, we get a lot of meaningful results. And this test has validated not only the feasibility of the new storage patten, but also the important value of link-base OLAP technology.