The Study On Multi-Relational Data Mining

Nowadays, most of the unstructured data is stored in relational database consisting of many tables. Numerous analysis and data mining tasks in a wide variety of applications including intelligence analysis, social network analysis, business data analysis, web data mining and bio informatics are based as much on the links among heterogeneous entities and events as the properties of individual entities. Hence, the databases in these applications contain both attribute and semantic relationship data. This data is stored in multi-relational (multi-table) form in relational database systems as a set of linked tables each corresponding to some conceptual entity or relationship. Multi-relational data mining (MRDM) is concerned with the discovery of models and patterns from such databases.Traditional data minig algorithms usually only focus on analyzing data organized in one single table, while multi-relational association rule mining is a process to find relationships between attributes from one single table or many tables by analyzing the data in the tables of a relational database.Multi-Relational Data Mining is the multi-disciplinary field dealing with knowledge discovery from relational databases consisting of multiple tables. Mining data which consists of conplex or structured objects also falls within the scope of this field, since the normalized representation of such objects in a relational database requires multiple tables. The field aims at integrating results from existing fields such as inductive logic programming, KDD, machine learing and relational databases, producing new techniques for mining multi-relational data and practicical applications of such techniques.This essay first introduces the function and model of traditional data mining, the processes of data mining and the application and perspective of data mining. Then, it introduces multi-relational data mining and its research meaning and range and its application situation. After that, it discusses classic algorithms for multi-relational data mining, including ILP, relational decision trees, relational distances based learning and etc., in detail. At last, it brings a new perspective of multi-relational data mining and improves it.This paper inproves the new perspective in multi-relational patterns mining: Iceberg-cube algorithm. The improved algorithm can handle a wider range of situations.The main works and achievements of this paper are:1. It discusses the concepts of data mining and introduces multi-relational data mining.2. It introduces the current situation in multi-relational data mining and discusses some classic algorithms to solve it.3. It proposes a new algorithm which is applicable to an ordinary database. The experiments show it has more efficiency and effectiveness.