MCTOPE Ensemble Machine Learning Framework: A Case Study of Routing Protocol Prediction

Many crucial applications of wireless sensor networks rely radically on routing protocols for an efficient data delivery. This paper presents a case study of scrutinizing the use-fulness of hybridization of machine learning classifiers in order to develop a Multi-Criteria Topsis based Ensemble (MCTOPE) framework. Technique for Order of preferences by similarity to Ideal Solution (TOPSIS), a multi-criteria assessment algorithm is employed to optimize the built ensemble learner for the prediction of an optimal reactive routing protocol for a wireless sensor network (WSN). The performance of the framework is first validated using six different machine learning datasets, and then the proposed method is implemented as a web application using R script and Python Django web framework. After experimenting with more than thousand combinations of training samples and ten base classifiers for the routing protocol prediction problem, MCTOPE framework builds an ensemble of support vector machine and neural network classifiers with an accuracy of 99.6%, which is far better, when it is compared with the performance of state-of-the-art classifiers. With the appearance of tremendous growth of machine learning classifiers in plenty of applications, an automatic ensemble building machine learning technique helps in minimizing the risk of obtaining poor results from a single classifier system, and will play a big part for efficient predictions in the future.