Title:Enhanced version of AdaBoostM1 with J48 Tree learning method

Abstract:Machine Learning focuses on the construction and study of systems that can learn from data. This is connected with the classification problem, which usually is what Machine Learning algorithms are designed to solve. When a machine learning method is used by people with no special expertise in machine learning, it is important that the method be robust in classification, in the sense that reasonable performance is obtained with minimal tuning of the problem at hand. Algorithms are evaluated based on how robust they can classify the given data. In this paper, we propose a quantifiable measure of robustness, and describe a particular learning method that is robust according to this measure in the context of classification problem. We proposed Adaptive Boosting (AdaBoostM1) with J48(C4.5 tree) as a base learner with tuning weight threshold (P) and number of iterations (I) for boosting algorithm. To benchmark the performance, we used the baseline classifier, AdaBoostM1 with Decision Stump as base learner without tuning parameters. By tuning parameters and using J48 as base learner, we are able to reduce the overall average error rate ratio (errorC/errorNB) from 2.4 to 0.9 for development sets of data and 2.1 to 1.2 for evaluation sets of data.