An Improved Adaboost Algorithm Based on Uncertain Functions

Abstract

Boosting is one of the algorithms which can boost the accuracy of weak classifiers, and Adaboost has been widely and successfully applied to classification, detection and data mining problems. In this paper, a new method of calculating parameters, Adaboost-AC, which uses the accelerated good fitness function to acquire the weights of the weak classifiers is presented. The new algorithm is compared with the tradition Adaboost based on the UCI database and its promising performance is shown by the experimental results.