•  
  •  
 

Document Type

Article

Keywords

Hybrid Model, Machine Learning, Anomaly detection, Cybersecurity, Threat Prediction, Time series analysis

Abstract

In today's digital era, cybersecurity has become a principal concern because of the increasing frequency and advancement of cyber threats. This study explores machine learning models for detecting and predicting anomalies in cybersecurity datasets. The research evaluates models such as linear regression, decision tree, RF, gradient boosting, KNN, SVR, LSTM, and neural networks utilizing performance metrics such as accuracy, MAE and MSE. A hybrid model that integrates different learning strategies is additionally proposed to improve the predictive accuracy and strength. The results highlight the superiority of ensemble approaches, especially the hybrid model, in improving peculiarity detection capabilities. The comparative analysis demonstrates that traditional models struggle with nonlinear patterns, whereas hybrid approaches successfully relieve this limitation. Moreover, this study emphasizes the importance of temporal data analysis for proactive threat detection and response. By leveraging diverse machine learning methods, this research contributes to strengthening cybersecurity infrastructures, enabling early threat detection, and minimizing security breaches. These discoveries emphasize the importance of adopting a comprehensive machine learning system to support cybersecurity resilience.

Download

Share

COinS