Abstract: In order to improve the prediction accuracy of protein secondary structure，exact radial basis function neural network and generalized regression neural network are presented.Multi-radial basis network predictors are established based on 5 encoding，Profile encoding，different sliding windows and cross validation method.These predictors are used to predict protein secondary structure，and it gets preferable results.Thereinto，aveQ3 is improved to 70.96%.The results not only show radial basis network models can increase the prediction accuracy efficiently，but also prove the validity and feasibility of these motheds.

Key words: radial basis neural network, protein secondary structure, predictor

摘要： 为提高蛋白质二级结构预测的精确度，提出并构建精确的径向基神经网络、广义回归神经网络，并基于5位编码和Profile编码，采用不同大小的滑动窗口，利用交叉检证法构建多个径向基网络预测器，分别对蛋白质二级结构进行预测，得到了较好的实验结果，其中aveQ3提高到70.96%。结果表明，径向基神经网络模型能有效提高预测精确度，也证明了实验方法的有效性和可行性。

关键词: 径向基神经网络, 蛋白质二级结构, 预测器