This paper focuses on the gas data processing and pattern recognition modules of detection and proposes a time-domain feature extraction method based on Singular Spectrum Analysis (SSATFE). This method effectively captures and reduces the dimensionality of gas signals, making the data more manageable for classification. To evaluate the extracted features, tree-based classifiers like decision trees and random forests were used. These models handle non-linear relationships well and are robust to noise, which is common in gas sensor data. Experiments on the UCL dataset show the classifiers’ ability to use the key components extracted by the SSA method for accurate gas detection.