Complexity Approximation of Classification Task for Large Dataset Ensemble Artificial Neural Networks

Abstract

In this paper, operational and complexity analysis model for ensemble Artificial Neural Networks (ANN) multiple classifiers are investigated. The main idea behind this, is lie on large dataset classification complexity and burden are to be moderated by using partitioning for parallel tasks and combining them to enhance the capability of a classifier. The complexity of the single ANN and ensemble ANN are obtained from the estimates of upper bounds of converged functional error with the partitioning of dataset. The estimates derived using Apriori method shows that the use of an ensemble ANN with different approach is feasible where such problem with a high number of inputs and classes can be solved with time complexity of ${\text{O}}\left( {{\text{n}}^{\text{k}} } \right)$ for some ${\text{k}}$, which is a type of polynomial. This result is in line with the importance of good performance achieved by diversity rule applied with the use of reordering technique. As a conclusion, an ensemble heterogeneous ANN classifier is practical and relevance to theoretical and experimental of combiners for ensemble ANN classifier systems for large dataset.