Abstract: In order to predict short fiber index of raw cotton，a BP neural networks model was designed on the basis of prediction method of the BP neural networks. Based on the raw cotton in southern Xinjiang，taking the seed cotton moisture regain and the rotational speed of saw cylinder as the basic characteristic quantity of BP neural networks models，the correlation between the short fiber index of raw cotton and the input parameters and the BP neural networks prediction model were proposed for prediction of the short fiber content. The results show that the BP neural networks model has strong ability for nonlinear approach which can actually reflect the nonlinear relationship between the short fiber index of raw cotton and main controlling factors. The small errors between the prediction values and measured values were achieved. The R-square of regression function between output and target of neural networks model for testing sample is 0.96357.