Increment learning method for continuous attribute measurement selection under C4.5 decision tree algorithm

Abstract

The invention discloses an increment learning method for continuous attribute measurement selection under a C4.5 decision tree algorithm. An increment learning process is formed by improving a continuous attribute measurement selection process in a C4.5 decision tree by use of a backward propagation algorithm in a neural network; and since the backward propagation algorithm is mainly applied to attribute selection, the method mainly studies the perspective of improvement of continuous attribute threshold selection. According to the scheme, an increment learning function is added compared to a conventional scheme, from the perspective of a continuous attribute threshold, recalculation of an original whole tree or a part of branches is abandoned, and such a complex and low-efficiency method, through combination of the high-efficiency learning scheme of backward propagation, increases functions of the C4.5 decision tree and is also a novel learning mode.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to an incremental learning method for continuous attribute measurement selection under the C4.5 decision tree algorithm. Background technique [0002] With the increase of the current amount of data, mining valuable information from the data has become a hot topic in current research. Data mining originated from KDD. In data mining, common and effective data mining algorithms include classification, clustering, association, and linear regression. etc., BI is an important application field. In 1984, a number of statisticians published the CART algorithm, which introduced the generation process of the binary decision tree. As a famous decision tree C4.5 algorithm in the classification algorithm, it has the reputation of one of the top ten algorithms for data mining. It can The data is presented in the form of a tree, and non-professional personnel can also make accurate judgm...

AI Technical Summary

Problems solved by technology

[0031] Purpose of the invention: Aiming at the problem that the C4.5 decision tree algorithm in the prior art (1) cannot incrementally learn for continuous attribute measurement selection (or if incremental learning is required, the example and the original training set can only be combined into The new training set can only be realized by using the C4.5 decision tree generation specification again), and select some branches in the prior art (two) to regenerate the C4.5 decision tree branch. The present invention provides a C4.5 The incremental learning method of continuous attribute measurement selection under the decision tree algorithm does not need to regenerate branches, only needs to change the threshold of continuous attributes and adjust the threshold in combination with the back propagation algorithm to achieve incremental learning, which can effectively reduce the number of iterations, Reduce computational complexity and improve the accuracy of C4.5 decision tree

Images