The invention discloses an epilepsy electroencephalogram (EEG) signal classification method based on fuzzy entropy, comprising performing fuzzy entropy analysis on attack intermission EEG signals and attack stage EEG signals; through characteristics, selecting and extracting fuzzy entropy of a reaction signal characteristic corresponding electrode as an input characteristic; and employing a screened fuzzy entropy of a corresponding electrode as an input characteristic to perform classification detection on epilepsy EEG signals. The method comprises characteristic calculation, characteristic extraction and classification modules, wherein characteristic calculation and extraction modules employ fuzzy entropy to analyze attack stage EEG signals and attack intermission EEG signals, and select an electrode with large KS detection difference as characteristic input; the classification module utilizes a support vector machine to perform classification detection on epilepsy EEG signals, and a support vector machine algorithm converts a problem into a dichotomy problem, thereby reducing calculating complexity; meanwhile, the method possesses good instantaneity, can be used for epilepsy detection and early warning, and has high sensitivity, singularity and accuracy.