A Novel Approach for Introducing Cluster Size Reduction and Diversity into an Optimized Ensemble Classifier PROJECT TITLE : A Novel Method for Creating an Optimized Ensemble Classifier by Introducing Cluster Size Reduction and Diversity ABSTRACT: Within the scope of this research project, a novel approach to generating an improved ensemble classifier is proposed. The issue of class imbalances is mitigated by the proposed method, which does this by partitioning the input data into its various data classes. After that, an incremental clustering process is used to generate a pool of class pure data clusters from the partitions. After the clusters of data have been generated, they are then balanced by adding samples from all classes that are located in the cluster centroid's immediate vicinity. In this way, all of the data clusters that are generated can be made to be balanced, and classifiers that are trained on such data clusters can also be made to be unbiased. This results in a more varied input space, which is necessary for the training of base classifiers. After that, the pool of clusters is put to use to educate a variety of different base classifiers, which ultimately results in the generation of the base classifier pool. After that, the pool of classifiers is approached as a combinatorial optimization problem, and an evolutionary algorithm is incorporated into the process. The strategy that has been proposed results in the generation of an improved ensemble classifier that not only has the potential to achieve the highest classification accuracy possible, but also has a smaller component size. The proposed method is validated using 31 benchmark datasets retrieved from the Machine Learning repository at UCI, and the results are analyzed and compared with those obtained by existing state-of-the-art ensemble classifiers. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest A Review of Deep Unsupervised Single-Source Visual Domain Adaptation Fast and Robust Representative Selection from Manifolds Using a Multi-Criteria Approach