PATTERN RECOGNITION LEARNING FOR BIG DATA IN CYBER SECURITY

Abstract

Digital security with regards to enormous information is known to be a basic issue and exhibits an extraordinary test to the exploration network. Machine-picking calculations have been recommended as a possibility for taking care of large information security issues. Among these algo-rithms, bolster vector machines (SVMs) have made astounding progress on different classification issues. In any case, to set up an effective SVM, the client needs to define the correct SVM configuration ahead of time, which is a difficult errand that requires master knowledge and a lot of manual effort for experimentation. In this work, we plan the SVM configuration process as a bi-target advancement issue in which exactness and model unpredictability are considered as two conflicting goals. We propose a novel hyper-heuristic structure for bi-target enhancement that is free of the issue area. This is the first time that a hyper-heuristic has been created for this issue. The proposed hyper-heuristic system comprises a significant-level methodology and low-level heuristics. The significant level system utilizes the inquiry execution to control the determination of which low-level heuristic ought to be utilized to create another SVM configuration. The low-level heuristics each utilize different rules to effectively investigate the SVM configuration search space. To address bi-target streamlining, the master presented structure adaptively incorporates the qualities of disintegration and Pareto-based ways to deal with inexact the Pareto set of SVM configurations. The effectiveness of the proposed system has been assessed on two digital security issues: Microsoft malware large information classification and abnormality interruption location. The obtained outcomes show that the proposed system is effective, if not unrivaled, compared to its partners and different calculations.