Based On The OCAT Logic Method Of Fine-grained Association Rules Mining Research And Application

With the wide use of computer technology, all walks of life has entered the information age, enterprise accumulated a large amount of historical data in the daily operations, at the same time, people are faced with historical data bloated but lack of useful information, how to make the fullest use of these huge amounts of data mining and effective information to support making decision, for each industry is very important. Generally speaking, the interactions between the data attributes to a certain influence on the transaction, such as temperature and humidity is how to affect the weather, rain or not related to what factors, etc., from a large amount of data and the data mining technology to analyze the relationship between the attribute, which can provide the decision support right from the side, so in has been extensively used in various industries. And intuitive association rule mining for simple description, expression and the advantages of the strong results can be interpreted as routine methods to discover this kind of correlation. Since put forward, through the perfect and improvement of many scholars, association rule mining algorithm has achieved rapid development, has been widely used in the production of life. In the actual demand, the association rules mining method for data mining from relational data set, usually with attributes for particle size to solve the correlation between the possible relevance attribute is not fine, to analyze the properties of the inner structure, the connotation and the hidden fine-grained related laws.Based on this, this article attempts to from the perspective of logic, in order to get through some algorithm mining the association rules of smaller size, for association rules mining provides a new Angle. The concrete realization of this paper is first based on the field characteristics in some binary encoding methods are put forward from the Angle of logic section properties on the properties of the decomposed, due to the conventional method of association rules mining and cannot handle size smaller segments of attributes, so we adopted is based on logic he (One Clause At A Time).He method requires data set can be divided into two categories, and can only and only suitable for dealing with binary data set, so you need to segments of the above property structure E+ and E- positive and negative training samples set. He method to get the process of association rules is the process of set covering, it time to get the current optimal child, the child can receive all samples in the E+, but refused to E-as much as possible in the sample, and when all child type or of conjunctive normal form to receive all the samples of E+, and accept all the negative samples, in E-algorithm is over, now get conjunctive normal form is the expression of association rules.In order to verify the validity of the method, the paper applies the method in the field of telecom customer churn prediction and meteorological rainfall forecast. The characteristics of the data set is different in different areas, so he logic method when applied to specific areas of need the light of the specific conditions of the data set, for example, the paper for the telecom customer churn prediction by the number of data aggregation class is more, so fewer bits to represent the classes; For rainfall forecast data aggregation class number is less, so you can according to each change of the binary number to represent the classes. The experimental results show that the particle size based on the method of association rules is reduced, the association rules of fine precision is improved, thus more intuitive, association rules is a form of conjunctive normal form at the same time, easy to implement parallel computing to improve efficiency, can well meet the needs of mass data mining.