Classifier integration method based on floating classification threshold

Abstract

The invention discloses a classifier integration method based on floating classification threshold, which is characterized by obtaining T optimal weak classifiers are by means of training after T iterations and then combining the T optimal weak classifiers to obtain an optimal combined classifier. In case of aiming at a bi-classification problem, training the T optimal weak classifiers comprises the steps of: (3.1) training the weak classifiers based on a training sample set S with weight omega<t>, wherein t is equal to 1,..., T; (3.2) based on the result of the step (3.1), adjusting sample weights omega<t+1>=omega<t>exp(-yiht(xi)) / Zt; (3.3) judging whether t is smaller than T, if so, enabling t to be equal to t + 1 and returning to the step (3.1) until t is equal to T; in case of aiming at multi-classification problem, training the T optimal weak classifiers comprises the steps of: (3.1) training the weak classifiers based on the training sample set S with weight omega<t>, wherein t is equal to 1,..., T; (3.2) based on the result of the step (3.1), adjusting sample weights shown in the description; (3.3) judging whether t is smaller than T, if so, enabling t to be equal to t + 1 and returning to the step (3.1) until t is equal to T. Compared with the prior art, the classifier integration method of the invention can overcome the defect that fixed classification threshold-based classifiers have unstable classification at points adjacent to classification boundary.

Description

technical field

[0001] The invention belongs to machine learning and pattern recognition methods, in particular to a classifier integration method based on a floating classification threshold to improve the performance of the classifier. Background technique

[0002] Improving classification accuracy through the combination of multiple classifiers has always been the main content of ensemble learning research, and the weak learning theorem strongly supports the feasibility of this research idea. Among them, AdaBoost (adaptive boosting, adaptive enhancement algorithm) and continuous AdaBoost algorithm based on the idea of ​​Boosting are currently one of the most researched and applied integrated learning algorithms, and their good performance and easy-to-use characteristics have attracted a large number of researchers. It is improved and perfected. Liu Dayou and others proposed a multi-classifier integration method based on incremental naive Bayesian network in patent CN1012...

Images