The success of AdaBoost and its application in portfolio management

We develop a novel approach to explain why AdaBoost is successful classifier. By introducing measure of the influence noise points (ION) in training data for binary classification problem, we prove that there strong connection between ION and test error. further identify decreases as iteration number or complexity base learners increases. confirm it impossible obtain consistent classifier without deep trees some complicated situations. apply portfolio management via empirical studies Chinese market, which corroborates our theoretical propositions.