Abstract: Feature extraction and the parameter optimization of classifiers are two key methods for the improvement of the classification accuracy.The paper uses normalization method for the feature transformation based on the public database UCI.KNN，PNN and SVM are employed for classification.The effects of normalization on the accuracy of classification and parameter optimization are discussed.The results of experiment show normalization improved effectively the accuracies of classifiers，especially for SVM，reduce the searching range of the parameters of classifiers and the training periods.

Key words: normalization, feature vector, parameter optimization, Support Vector Machine（SVM）

摘要： 特征提取和分类器的参数优化是提高分类准确率的主要途径，对公用数据库UCI的相关数据进行特征向量的归一化处理，采用KNN、PNN和SVM进行分类。讨论了特征归一化对分类准确率和分类器参数的影响。实验结果表明：归一化能有效提高分类器的分类准确率，SVM尤为明显，且参数的寻优范围缩小，缩短训练周期。

关键词: 归一化, 特征向量, 参数优化, 支持向量机