A Simple Ensemble Learning Algorithm For a Real Time High Dimensional Data

Abstract

High dimension data is growing rapidly, which challenges the existing data mining and learning systems. A stable performance can be achieved precisely in high dimension data using machine learning algorithm, ensemble learning and feature selection methods. The ensemble classifiers like Decision tree, Random forest, support vector machine, Navies Bayes, Linear Discriminant Analysis are applied as a self-contained module and the output label is reliably produced by the fusion of majority voting for data analysis. At present a lot of investigations for reduction of high dimensional data set are implemented, but the study result shows only a minor improvement in accuracy. In this paper, a wrapper method is properly used to improve certain key features such as random state, tune max depth, tune number samples split, and tuned loss function. The proposed work is carried out using real-time garbage data set collected from the municipal corporation of Tiruchirappalli, India and the selected sample is tested with FGDT (Forward Gradient Boost Decision Tree) shows improvement in accuracy.