Matrix increment reduction method based on knowledge granularity

Abstract

The invention discloses a matrix increment reduction method based on the knowledge granularity. The matrix increment reduction method includes the following steps: firstly, REDU of an existing decision table (called as a former decision table) is calculated through an equivalence relation matrix; secondly, after some new objects are added into the former decision table, a new decision table is obtained, the knowledge granularity of the REDU in the former decision table and the knowledge granularity of the new decision table are calculated with the matrix increment method, whether the knowledge granularity of the REDU in the former decision table is equal to the knowledge granularity of the new decision table or not is judged, if the knowledge granularity of the REDU in the former decision table is not equal to the knowledge granularity of the new decision table, the external importance of all attributes except for the REDU is calculated in the new decision table with the matrix increment method, the attributes with the highest external importance are sequentially selected in a circulating mode to be added into the REDU, and then the knowledge granularity of the REDU is calculated till the knowledge granularity of the REDU is equal to the knowledge granularity of the new decision table; finally, the redundancy attributes in the REDU are cancelled in a circulating mode, and the REDU of the new decision table is obtained. By means of the matrix increment reduction method, the REDU is effectively and rapidly solved when the dynamic objects in the decision table are increased, and therefore the knowledge discovery efficiency is easily improved.

Description

technical field [0001] The invention relates to the technical field of granular computing and rough set theory in artificial intelligence, in particular to a matrix incremental reduction method based on knowledge granularity. Background technique [0002] At present, with the continuous development and wide application of storage technology, information systems in practical applications not only accumulate a large amount of various data, but also these real-time data will change every moment. It takes a lot of time to mine complex data, which makes it difficult for batch learning algorithms to extract useful information and knowledge from dynamic data in a timely and efficient manner. Incremental learning is a learning system that can continuously learn new knowledge from new samples from the environment and retain most of the previous knowledge. By using incremental learning technology, knowledge can be updated gradually, and previous knowledge can be corrected and strengt...

AI Technical Summary

Problems solved by technology

[0002] At present, with the continuous development and wide application of storage technology, information systems in practical applications not only accumulate a large amount of various data, but also these real-time data will change every moment. It takes a lot of time to mine complex data, which makes it difficult for batch learning algorithms to extract useful information and knowledge from dynamic data in a timely and efficient manner.

Images