Adaptive multi-classifier fusion approach for gene expression dataset based on probabilistic theory

The existing focal point of research in pattern recognition is ensemble learning which is also known as classifier fusion. The classifications of algorithms are used to assemble the different types of models for augmenting the performance. In this paper, we propose a novel fusion-procedure which has used Naïve Bayesian classifier that acts as a supervisor to optimize the parameters and enhance the performance. The classifiers such as; Particle Swarm Optimization-Functional Link Neural Network (PSO-FLANN), Bat inspired-Functional Link Artificial Neural Network (BAT-FLANN) and Support Vector Machine (SVM), used as base classifiers are diverse to each other. The kk-fold cross validation is used for training and testing of the datasets. The performance of the model has been compared with recent classifier fusion techniques such as Uniform Voting, Distribution Summation, Dempster–Shafer, Entropy Weighting and Density based Weighting on six benchmark gene expression datasets under the objective functions like accuracy, parallel execution, time complexity and space complexity. A greater competitive accuracy has been achieved in comparison with the performance of other fusion strategies, aiming at our goal-functions.