为比较各种混沌神经网络预测算法的误差性能 ,该文提出了预测误差评价准则 ,即根均方误差 ,偏差 ,预测精度 ,决定度系数 ,绝对误差 ,以及一些归一化根均方误差等 ,并分析了它们在描述误差特征上的具体含义。针对两种混沌神经网络预测算法 (即全局神经网络算法和局部神经网络算法 ) ,利用该准则进行了性能分析 ,给出了合理的评价。结果表明 ,与混沌神经网络预测的局部模型算法相比 ,全局模型算法有更好的预测效果 ,且训练时间短 ,占用资源少 ,推广能力好 In order to compare the error performance of various chaotic neural network prediction algorithms, this paper presents the evaluation criteria of prediction error, such as root mean square error, bias, prediction accuracy, coefficient of decision, absolute error, and some normalized root mean square error , And analyzed their specific meaning in describing the error characteristics. Aiming at two kinds of chaotic neural network prediction algorithms (ie, global neural network algorithm and local neural network algorithm), this criterion is used to analyze the performance and gives a reasonable evaluation. The results show that, compared with the local model algorithm of chaotic neural network prediction, the global model algorithm has better prediction effect, and the training time is short, less resource consumption, good promotion ability