中文摘要：

采用WEKA特征选择法分析了原棉的性能指标对成纱质量的影响，并结合BP人工神经网络对成纱条干、成纱强度和成纱强度不匀进行了预测。采用WEKA特征选择法可以减少BP神经网络的输入节点数，与单纯的BP神经网络的预测结果相比，WEKA特征选择法结合BP神经网络预测结果较准确，预测值与实测值之间的平均相对误差较小。

英文摘要：

In this paper, the effect of cotton fiber property on formed yam quality is analyzed by WEKA feature selection. Combined with BP neural network, formed yarn evenness, strength and strength irregularity are predicted. The WEKA feature selection is expected to reduce the input layer node numbers of BP neural network. Compared with that of pure BP neural network, the result got from the combined one is more precise; the relative mean error between the measured values and the forcast results of cotton yarn quality is reduced.