Adaptive hybrid optimization enabled stack autoencoder-based MapReduce framework for big data classification

The rapid development in the field of network and technology has led to the creation of massive digital information. The main problem found in the big data context is the classification of the task. In this paper, a big data classification model based on the optimization-enabled MapReduce framework was developed for the effective management of data. The optimization algorithm, namely the adaptive Exponential Bat algorithm (adaptive E-Bat) is used, which is the integration of the adaptive concept along with the E-Bat algorithm. The entire operation for big data classification is briefed as follows: The data from the various sources is provided to the Mapper, which performs feature selection. The mapper has the adaptive E-Bat algorithm for selecting the suitable features from the database, where the adaptive nature of the algorithm facilitates the effective handling of the real-time data. Then, the selected features are provided to the reducer for actual big data classification. The reducer uses a deep stack autoencoder, which is trained using the adaptive E-Bat algorithm. Thus, the big data classification is facilitated using the proposed adaptive E- Bat-based stack autoencoder. The performance is analyzed in terms of accuracy and TPR. The proposed method outperformed the existing methods with the TPR and accuracy of 0.9519 and 0.8784, respectively.