Adaptive Exponential Bat algorithm and deep learning for big data classification

The concept of big data refers to the huge amount of information that the organizations process, analyse and store. In the real-world scenario, some big data possess other features such as credit card fraud detection big data, extreme weather forecast big data and so on. In order to deal with the problem of classifying the binary imbalanced big data, based on MapReduce framework (MRF), an enhanced model is proposed for the process of classification in this paper. An optimization based on MRF is used for dealing with the imbalanced big data using the deep learning network for classification. The mappers in the MRF carry out the feature selection process with the proposed Adaptive E-Bat algorithm, which is a combination of adaptive, Exponential Weighted Moving Average (EWMA) and the Bat algorithm (BA) concepts. Using the features, the reducers perform the classification using Deep Belief Network (DBN) that is trained with the proposed Adaptive E-Bat algorithm. The performance of the proposed Adaptive E-Bat DBN method is evaluated in terms of metrics, namely accuracy and True Positive rate (TPR); a higher accuracy of 0.8998 and higher TPR of 0.9144 are obtained, that show the superiority of the proposed Adaptive E-Bat DBN method in effective big data classification.