Ensemble model with combined feature set for Big data classification in IoT scenario

Abstract

Sensor nodes that are wirelessly connected to the internet and several systems make up the Internet of Things system. Large volumes of data are often stored in big data, which complicates the classification process. There are many Big data classification strategies in use, but the main issues are the management of secure information as well as computational time. This paper's goal is to suggest a novel classification system for big data in Internet of Things networks that operates in four main phases. Particularly, the healthcare data is considered as the Big data perspective to solve the classification problem. Since the healthcare Big data is the revolutionary tool in this industry, it is becoming the most vital point of patient-centric care. Different data sources are aggregated in this Big data healthcare ecosystem. The first stage is data acquisition which takes place via Internet of Things through sensors. The second stage is improved DSig normalization for input data preprocessing. The third stage is MapReduce framework-based feature extraction for handling the Big data. This extract features like raw data, mutual information, information gain, and improved Renyi entropy. Finally, the fourth stage is an ensemble disease classification model by the combination of Recurrent Neural Network, Neural Network, and Improved Support Vector Machine for predicting normal and abnormal diseases. The suggested work is implemented by the Python tool, and the effectiveness, specificity, sensitivity, precision, and other factors of the results are assessed. The proposed ensemble model achieves superior precision of 0.9573 for the training rate of 90 % when compared to the traditional models.