The statistical analysis and artificial neural networks were applied to the prediction of cotton yarn strength with fiber properties measured by an HVI instrument. The linear and non linear relationships between fiber properties and yarn strength were compared in the multiple regression analysis. The coefficients of correlation and determination as well as the adjusted coefficients of determination were calculated in order to select effective parameters among the various fiber properties. Neural networks with the back propagation learning algorithm were trained to match a set of input data with the corresponding set of output data, and their weights were obtained. The results obtained by the statistical analysis and neural networks were compared with each other in terms of the mean of square error between the target values and the experimental ones. It is therefore concluded that neural networks provide an effective tool alternative to the statistical method for yarn strength prediction and do not require a specific model necessary for the regression method.