Cross-sensitivity of gas sensor array severely affects the measurement precision.Back propagation（BP）

radial basis function（RBF） and fuzzy neural networks（FNN） have been built in MATLAB neural networks toolbox and applied in predicting volume fraction of gas mixture（formaldehyde

toluene

acetone and ethanol）.The results show that the precision of FNN is higher than BP and RBF on volume fraction prediction of gas mixture.Moreover

data pre-processing by PCA and ICA have been proved to be helpful on raising prediction precision.