Gas identification technology, essential in diverse sectors such as industrial production, environmental monitoring, food safety, and medical diagnosis, leverages gas sensors and data processing for precise analysis. This study focuses on machine learning in gas recognition, detailing the use of feature extraction techniques and algorithm applications in conventional methods and comparing different algorithms. It further assesses the effectiveness and characteristics of convolutional neural networks and recurrent neural networks in deep learning for gas recognition, while exploring optimization strategies like multi-task learning and attention mechanisms. The paper concludes by addressing the current challenges in gas recognition technologies and outlining prospective research areas.