In this paper, the prediction model of the nearest adjacent yarn knot and the BP neural network prediction model are used to predict the cotton knot index of fine yarn varieties, and the prediction effect of the two prediction methods is analyzed according to the average relative error of the prediction data set and the average square root error.