In order to analyze the freshness of fish meal comprehensively and quickly, the electronic nose and SPME-GC-MS were combined to evaluate the freshness of fish meal through the content of volatile components. The volatile components of fish meal with 18 different storage times levels was analyzed by solid-phase headspace microextraction, and the kinds and contents of volatile components in fish meal were obtained at different storage time, and the change rule of volatile components was obtained. Multiple linear regression and KNN regression were used to model various volatile components and freshness markers of fish meal and the response of electronic nose sensors. A total of 101 volatile compounds were detected, including 11 alcohols, 20 aldehydes, 26 ketones, 8 acids, 9 hydrocarbons, 15 nitrogen-containing compounds, 4 sulfur-containing compounds and other compounds. Among them, there were more aldehyde ketones, less esters, phenols and ethers. Ethoxyquin and γ-dehydrogenase were also detected in the volatile gas of fish meal, the reason was that fish meal, as a feed material, was not only a simple processing of raw fish, and also required other additives to facilitate the transportation and storage of fish meal. At the beginning of storage, the relative content of aldehydes was 48.99%, but at the end of storage, the relative content of aldehydes was 0.95%. The relative content of aldehydes was significantly reduced, while the relative content of ketones was increased. At the beginning of storage, the relative content of ketones was 28.55%, but at the end of storage, it was 51.43 %. The results showed that the freshness of fish meal was characterized by trimethylamine, 3-methyl-1-butanol, dimethyl disulfide and dimethyl trisulfide. The relative contents of alcohols showed a downward trend after the first increase, acids, nitrogenous compounds and sulfur compounds showed an upward trend. The models of various volatile components in fish meal and the markers representing freshness and the sensor responses of electronic nose were established by multiple linear regression method and KNN regression method. The results showed that the KNN regression method had higher accuracy than multiple linear regression method. The correlation coefficient was between 0.7633 and 0.9999, and RMSE was between 0.0867% and 8.4655%.Therefore, these volatile components can be well predicted from electronic nose measurement. The results can be used as reference for understanding the flavor composition and change rule of volatile components of fish meal during storage, and for judging the freshness of fish meal according to the smell component.