Information retrieval using mutual refinement and deep learning neural network techniques in big data

Abstract

Big data analytics is popularly enhancing to attain valuable information that can be used enormously in scientific and business applications. The retrieval of information provides a better analysis in many areas including education, business, security and health care. The big data includes various data types such as structured, unstructured and semicrucial role which provides different kind of solutions as per the user request. Numerous research works have been designed to retrieve the information from the big data with enhanced optimization techniques. Due to huge volume and variety of data, accessing the information from big data becomes a difficult task. It leads to the cause of data complexity with higher retrieval time and also fails to retrieve the information from the big dataset effectively. The proposed research work overcomes these issues by using three different techniques such as Mutual Refinement Technique (MRT), Deep Multi-Label Hashing technique (DMLH) and Deep Learning Modified Neural Network technique (DLMNN). The proposed techniques have been developed for achieving enhanced information retrieval from big data with reduced computation, retrieval time with high accuracy. The performance of information retrieval from big data has been improved by proposing a Mutual Refinement Technique. The main aim of MRT is to retrieve the information from big data with minimum retrieval time along with high accuracy. Here, data visualization has been achieved through the training phase and retrieval phase. During training phase the education big dataset Open Government Data (OGD) platform - India for the academic year (2010-17) is considered as input big dataset. It includes varietyof text documents such as CSV, XLS, XLSX and JSON. From the education big dataset, data preprocessing such as stemming, stop word removal and removal of redundant data has been carried out to remove unwanted data and to split data into various tokens. After preprocessing, feature extraction is implemented to extract features such as