The invention relates to an electronic nose apparatus evaluating method for the sensory quality of tobacco and a tobacco product. The method is characterized in that: a breathing process of a professional is simulated, and a procedure of detecting and evaluating the tobacco and tobacco product samples by a tobacco electronic nose apparatus is determined; a large quantity of standard samples are tested to establish tobacco big data of the relationship between the gas sensor array response and the brand mark and sensory quality index evaluation; the issue of tobacco sensory quality evaluation isconverted into an issue of predicting recognition and sensory quality indexes, and is solved by a modular neural network cascade model; an effective task decomposition method, a corrected S-shaped activation function and an optimized single-output single-hidden layer neural network structure are proposed to realize big data-oriented neural network fast learning; a majority vote decision rule anda quantification rule based on a plurality of neural network modules are proposed to realize recognition of large-scale tobacco and tobacco products as well as prediction of sensory quality index.