Research On Ensemble Learning Method And Application Based On Hierarchical Clusteirng

Ensemble learning is one of the four main research directions in machinelearning areas. Ensemble method is learning algorithms that constructs a set ofclassifiers and then classifies new data points by taking a (wset eighted) voteof their predictions. With the rapid development of the data collection and thedata storage technology, the scale of data for machine learning is increasing.Using a single learner can not resolve some learning issues properly. In orderto improve the generalization capability of learning systems, the method ofutilizing multiple learners to solve problems is proposed. Ensemble learning isprimarily used to improve the performance of a model, or reduce thelikelihood of an unfortunate selection of a poor. Nowadays, ensemble learningis widely used in sensor fault tolerance, handwritten character recognition,bio-certified, radiation source identification, linguistics, transportation,medicine and management filed.This paper introduces the ensemble learning algorithm and its appliedrange.The research work is as follows:(1) This paper gives a simple explanation on ensemble learning. Thenhave a detailed analysis on the problem of data size, calculation andclassification function description. The analysis results is proved byexperimental finally.(2) The paper puts forward a classifier ensemble algorithm that based onhierarchical clustering, For ensemble learning, it has long been the focushoping to improve the final performance through improving diversity amongbase classifiers. It is need to request the scale of ensemble learning in practicalwork application. Based on the analysis of the relationship between the overallperformance, the member classifiers performance and diversity.This algorithmuses the paired diversity as the clustering distance, and the accuracy as theselection criteria, the weak classifiers are selected for Boosting clustering byhierarchical clustering, it can significantly improve the diversity andclassification performance of the original system.(3)Using haar features and the improved Boosting algorithmimplements a multi-cascade classifier, and uses it for face detection. Theexperiment shows that multi-cascade classifier can achieve better recognition performance. Finally, using the improved face detection algorithm achievesan application system of face detection.