KDD on the CIFAR10 dataset by meta models for binary classification approaches through IP

Abstract

An ensemble classifiers have been successfully applied to image classification problems. This research finds the efficiency of ensemble model on tiny images for real world problems based on the following metrics like accuracy, precision, recall, receiver operating characteristic curve (ROC), precision recall curve (PRC), kappa, F-Measure, MCC, performance analysis of deviations on like Bagging, Logit Boost (LB), Iterative Classifier Optimizer (ICO), Classification Via Regression (CVR), Multi Class Classifier Updateable (MCCU), and Random Committee (RC) of selected meta models. The RC gives best outcome compare with other models such as 90% of accuracy, 90% of precision, 0.90 of recall, 0.96 of ROC, 0.94 of PRC, 0.80 of kappa, 0.82 of MCC with low deviations. The ICO gives worst outcome compare with other models such as 65% of accuracy, 0.65 of precision, 0.65 of recall, 0.74 of ROC, and 0.74 of PRC with high deviations. This work finds and evolves that the Random Committee model performs well based on comparison of several metrics with other models.