Hybrid Classifier Based on Binary Neural Network and Fuzzy Ant Colony Optimization Algorithm

Abstract

Since the last decade, classification methods are useful in a wide range of applications. Classification is a task to segregate the sample into different groups. This capability can be introduced in the computer system by designing various types of classifiers. The methodology known as the neural network in artificial intelligence (AI) offers so many approaches for classification problem. In this aspect, the hybrid classifier is created and tested for the application of the multi-class classification problem on the basis of Binary Neural Network (BNN) and Fuzzy Ant Colony Optimization (FACO). ACO algorithm has proven to be a powerful tool for optimal control of continuous-state dynamic systems. This hybrid Fuzzy ACO algorithm combines the multi-agent optimization and fuzzy partitioning of the state space of the system. This classifier is used to classify any type of data whether it is labelled or unlabeled or both. FACO can monitor the continuous data type and eliminate the complexity and allowing it to outperform the standard algorithm. FACO employs a rule-based feature selection concept to eliminate the least influential features, which increases computational time. To begin, it preprocesses the data set to generate binary values. Then, preprocessed data is used as an input for the hidden layer in order to reduce preparation time across all multiple classes. The hidden layer training is performed in parallel, which employs the geometrical learning principle. This system has been evaluated on a variety of benchmark datasets and compared by varying the learning parameters.