The increasing complexity and frequency of cyberattacks have made traditional rule-based detection systems insufficient for securing modern digital infrastructures. This research proposes a hybrid machine learning (ML) framework designed to enhance proactive threat detection and prevention capabilities in cybersecurity systems. The proposed model integrates intelligent feature selection techniques with a combination of classical ML algorithms and deep learning architectures to achieve higher accuracy, reduced false positives, and better adaptability to evolving threats. The study begins with a comprehensive analysis of benchmark cybersecurity datasets such as CIC-IDS2017, NSL-KDD, and UNSW-NB15. Various preprocessing steps are implemented, including data normalization, encoding, and class balancing. Feature selection methods such as Recursive Feature Elimination (RFE), Mutual Information, and Principal Component Analysis (PCA) are applied to identify the most relevant attributes contributing to threat prediction. The hybrid framework incorporates ensemble-based classifiers like Random Forest and Gradient Boosting, alongside deep learning models such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. Performance metrics including accuracy, precision, recall, F1-score, and AUC-ROC are used for evaluation. Results demonstrate that the hybrid approach significantly outperforms standalone models in terms of detection accuracy and generalization. This research contributes a scalable and adaptable threat detection solution suitable for real-time cybersecurity applications. The intelligent fusion of feature selection and hybrid learning models provides a robust defense mechanism against a wide spectrum of cyber threats, establishing a foundation for future advancements in automated and intelligent cybersecurity systems.