Secure Rule Generated and Predicted Deep Computational Model for Big Data Feature Learning on Cloud

Abstract

In big data analytic and mining feature learning is nothing but a basic and important topic. Numerous tasks are formulated on feature learning by the features of big data such as volume, variety as well as velocity those are related to huge amount of information, several types of information as well as the streaming speed of data, accordingly. Concentrations on these issues, a model of deep computations are implemented for big data feature learning. At the start, for the purpose of learning the features of heterogeneous data a basic deep computation model is outlined by modifying the deep learning model for the space of vector with the space of tensor . Next, for real time feature learning of big data, an incremental deep computation method having two incremental tensor auto-encoders is designed. At last, for maximizing the efficiency to train the parameters of deep computation model by using the cloud servers the privacy preserving technique is developed. In given system a privacy preserving high order back- propagation algorithm is developed by consolidating with the full homomorphic encryption, BVG for protection of the information which may b private while doing the deep computation model on cloud. Also in contribution we developed a rule generation method for creating the frequent pat- terns by making use of outcomes which are predicted by utilizing FP-Growth algorithm. Few of the test outcomes demonstrated that the implemented deep developed deep computation model can study features for big data efficiently as well as effectively. Majorly the last system is scalable by making use of many cloud servers that is specifically suitable for big data .