ABSTRACT
Optimization of sensor array is a significant topic in the application of electronic nose (EN). Stepwise discriminant analysis and cluster analysis combining with screening of typical index were employed to optimize the original array in the classification of 100 samples from 10 kinds of traditional Chinese medicine based on alpha-FOX3000 EN. And the identification ability was evaluated by three algorithm including principle component analysis, Fisher discriminant analysis and random forest. The results showed that the identification ability of EN was improved since not only the effective information was maintained but also the redundant one was eliminated by the optimized array. The optimized method was eventually established, it was accurate and efficient. And the optimized array was built up, that is, S1, S2, S5, S6, S8, S12.