The rapid development of cyber-attacks has emphasized the importance for advanced Intrusion Detection Systems (IDS) capable of recognizing and preventing sophisticated attacks. Conventional IDS approaches, often struggle with the complexity and variety of modern intrusions requiring the creation of more robust and durable solutions. This survey offers a comprehensive analysis of several Machine Learning (ML) techniques used in the creation of IDS, focusing on their methodologies, advantages, and potential drawbacks. The review explores the use of supervised, unsupervised and hybrid ML algorithms for intrusion detection, analyzing how well they detect intrusions in various scenarios and datasets. Furthermore, the paper explores influence of feature selection techniques on performance outcomes ML-based IDS. To determine the effectiveness of these methods over several datasets and situations, a number of performance metrics are investigated. In order to help academics and practitioners choose and optimize machine learning methods for reliable intrusion detection, this study attempts to give a thorough overview.