An electronic nose (e-nose), containing 16 tin metal oxide sensors with various sensitivities, was used to classify decay times in an 18 h accelerated decay study of tilapia (Oreochromis niloticus). Data collected were split into three 6 h base classes for training. Principal component analysis was tested for feature extraction to be used in classification but was found to be inadequate. Linear discriminate analysis was also used and found adequate for feature extraction. Both least squares and K-nearest neighbor classifiers were explored. Least squares and K-nearest neighbor produced classification rates of 86.4% and 87.0%, respectively. Data combing techniques were used to increase classification rates from 87.0% to 97.8% for K-nearest neighbor. Optimum classification performance was achieved with classes corresponding to 0-1.9 h, 6-7.9 h, and 12-13.9 h. The dataset was also classified into six 3 h classes. Data classifications for the 3 h classes followed trends expected for decaying freshwater fish. Data combing was again employed to increase the classification that was possible. A final classification was achieved of 78.8% for least squares and 83.8% for K-nearest neighbor.