Mining method for fuzzy rough monotonic data based on interval average

Abstract

The invention refers to the theory of fuzzy rough set and provides a mining method for fuzzy rough monotonic data based on interval average. The method includes: realigning decision properties and condition attributes; dividing realigned collections into intervals; deciding monotone according to each interval average; determining membership function values of the condition attributes; determining number of divisions according to circular division of the intervals to obtain function range of interference factors; setting filtering rules to filter unsuitable data so as to obtain reduction data collection and optimal data.

Description

technical field [0001] The invention relates to the technical field of data mining, in particular to design a fuzzy rough monotone data mining method based on an interval average value. Background technique [0002] Rough set theory is a mathematical tool used to deal with uncertain and incomplete data information, and fuzzy set can also describe the uncertainty of information and knowledge. Since the two are highly complementary, they can be combined to analyze information. Deal with uncertainty. In the decision table, rough set mining, the dependency relationship between conditional attributes and decision attributes, reducing attributes, finding out which conditional attributes are more important to decision attributes, the main theoretical basis is the equivalence relationship, due to the limitations of the equivalence relationship Many people have proposed different reduction relations, T.Y.Lin et al. proposed domain and compatibility relations, S.Greco et al. proposed...

AI Technical Summary

Problems solved by technology

[0003] In 1997 and 1998, C.J.Wu and Te-Shun Chou respectively introduced and discussed the fuzzy monotone function and its application in logic control. Some documents discussed the theory related to fuzzy monotone in the Mamdani-Assilians model and T-S inference method. In recent years Many people have discussed many algorithms for decision table attribute reduction, etc. In the decision table, assuming that the increase and decrease of the decision attribute quantity depends on the increase and decrease of some conditional attribute quantities, then it is necessary to mine the decision attribute quantity Change the conditional attribute that has an important impact. It is said that there is an important monotonic dependency between such a decision attribute and the conditional attribute, and this monotonic dependency is not necessarily strictly monotonic in the decision table, that is to say, the condition of two adjacent points The monotonicity of the attribute value does not necessarily map to the monotonicity of the corresponding two points of the decision attribute, because there are various interference factors and errors in the actual data, but the existing technology has not been able to effectively mine out the influence of the change of the decision attribute. The conditional attributes of important influences, and can affect the decision attribute by controlling these conditional attributes

[0004] The existing technical models are mainly expanded and changed around the equivalence relationship. Therefore, there are some problems when using these technical models for knowledge reduction and data mining. The details are summarized as follows: (1) In the face of many inputs and Outputting attributes and complex and huge data, how to construct an equivalence relationship between attribute data and some existing extended relationships is a relatively difficult problem; (2) decision tables composed of complex data are generally inconsistent decision tables, and Existing attribute reduction algorithms are generally based on consistent decision tables; (3) Data in complex environments are generally continuous data, and existing attribute reduction algorithms generally have to discretize continuous data , but for irregular complex, variable and large amounts of data, this is a difficult problem; (4) For the existing heuristic knowledge reduction methods, most of the core attributes are used as the starting point, and the relatively important largest attribute is given priority in each step It is required to reduce the result, but because of the problem raised in (1), it is not easy to obtain the core attribute, and it is also difficult to obtain the relative importance in a complex environment, because among the many attributes, the importance of the attribute is not It is easy to observe, and the input and output data are very complex, it is difficult to obtain the relative importance of attributes through artificial statistics or through existing analysis methods; (5) Since the data in complex environments are basically incomplete, This is a difficult problem for the existing attribute reduction methods; (6) The existing attribute reduction algorithms are generally aimed at limited data value sets, and are not suitable for a large number of irregular data value sets. The output data is often a large number of irregular data sets

Images