本文利用LibSVM软件包对已通过改进过的CKSAAP方法特征提取出来的数值特征对磷酸化位点进行预测。结果表明,基于SVM和CKSAAP结合氨基酸频率方法的水稻蛋白质磷酸化位点预测在丝氨酸、苏氨酸和酪氨酸的平均预测准确性为80.877%,马修斯系数为0.616。对比单纯使用CKSAAP准确性显著提高。与PlantPhos和Musite的预测性能的对比结果显示,在磷酸化酪氨酸位点的预测性能显著高于PlantPhos及Musite。 In this paper, we use the LibSVM software package to predict the phosphorylation sites by using the numerical features extracted from the improved CKSAAP method. The results showed that the prediction accuracy of serine, threonine and tyrosine was 80.877% based on SVM and CKSAAP combined amino acid frequency method, and the Mathews coefficient was 0.616. The accuracy of using CKSAAP alone was significantly improved. Comparisons with the predicted performance of PlantPhos and Musite show that the predicted performance at the phosphorylated tyrosine site is significantly higher than that of PlantPhos and Musite.