The Research Of Multi-Realation Assoaction Rules Algorithms In Data Mining

Learning method in traditional data mining algorithms, which is proposition algorithm, is attribute-value algorithm, each sample is expressed in the form of attribute-value group. With this representation, the attribute type is fixed, each attribute has a corresponding value given, so the data set can be seen as a relational database table or a relation. Each table row corresponds to a sample, and each column corresponds to a attribute. The fact is, the relational database in order to effectively organize and access data, the data structure in a complex form, and form relationships with many relations. The structure of relational databases expressed in different relations link between tuples, and this link shows the problem domain and some important background information on the structure of the sample content. Attribute-value of learning method to adapt only a single table can not be directly linked and the implied use of such content, and can not find the hidden form in the database. Although in theory more relational tables can be turned into a single table, but in fact there are many practical problems can not be avoided.Inductive Logic Programming (abbreviated as ILP) is cross areas of machine learning and logic programming. ILP-based multi-relational data mining, relational database devoted to the discovery of complex models involving multiple relations. Multi-relational data mining relational tables can be directly analyzed in a number of data without having to convert to a single data table. Because of its mode of representation language using the language of first order predicate logic, and propositional logic more expressive than the language, so it can express more complex models and to facilitate the use of background (field of) knowledge, and applications are more extensive, based on ILP relational data mining become a research hotspot today.This paper compares the system and complete description and analysis of the multi-relational association rule mining the theoretical basis and research, mainly the following:①Based on previous work and relevant literature, reviewed the concept of multi-relational association rules related to the nature, mining method, the main problem facing;②Studied the classic data mining algorithm Apriori and FP-growth, and according to the literature, combined with research, analysis of the proposed algorithm, characteristics and algorithm steps, and functions were analyzed.③Research the definition of the ILP technique, formal description and theoretical basis of PROLOG language and operating mechanism are .④Analysis of representative multi-relational association rules algorithm WARMR and FARMER, and analyze its advantages and disadvantages, and briefly describes WARMR expansion algorithm WARMeR.⑤The experiment platform, the classical Apriori algorithm was implemented and proposed improvement measures, achieved through PROLOG facts and rules of inference, analysis of the basis FARMER Algorithm,and implement the WARMR Algorithm in prolog ,mine the MRAR.