Mud logging system presents the most promising tool for exploring oil and gas resources, which relies on the multi-component analysis of gases to evaluate the hydrocarbon reservoirs. Higher detection efficiency, especially relating to the high-throughput analysis, can offer an important means to enhance the oil and gas recovery of hydrocarbon reservoirs. However, fewer technologies can measure both hydrocarbon and non-hydrocarbon gases in a timely manner. In the present work, a method based on data driven Raman spectroscopy (DDRS), which allows simultaneously accurate analysis of 12 hydrocarbon and non-hydrocarbon gases, has been proved. By developing a data driven strategy that incorporates higher-density discrete wavelet transform (HDWT) and template oriented frog algorithm (TOFA), the spectral resolution of Raman spectroscopy was significantly improved in both time and frequency domains to extract the target gas information from the Raman spectra, facilitating the detection of hydrocarbon reservoir. Theoretically, it is considered that the DDRS represents a promising tool for extracting the quantitative information from multi-component gases in the presence of dominantly uncontrolled variance. Our work has tested and refined the DDRS by performing a comprehensive analysis of three gas mixtures established to yield challenges representative of those encountered in a common mud logging analysis. The DDRS compared favorably with the gas chromatography (GC) method. The satisfactory results indicated that the DDRS has the capability to continuously detect 12 hydrocarbon and non-hydrocarbon gases in real time, which could be extended to online multi-component analysis of any other gas and fuel systems. (C) 2017 Elsevier Ltd. All rights reserved.