基于粗糙集和神经网络在小麦病害诊断中的应用

Application for Agricultural Pests Diagnosis Based on Rough Set and Neural Network

针对小麦病害诊断过程的复杂性以及处理信息的不确定性,综合了粗糙集理论与BP神经网络的各自优势,构建了小麦病害诊断模型。首先是对连续的样本数据进行离散化,主要采用差别矩阵计算方法进行启发式知识约简,得到最小简化规则,然后把约简结果作为BP神经网络的输入结点。实验结果表明,采用该方法不仅优化了神经网络的拓扑结构,还降低了神经网络的训练时间,同时大大提高了学习速度和故障诊断的准确率。

For wheat,the process of disease diagnosis and the complexity of the process of dealing with the uncertainty of information,the paper is constructed based on rough set(RS) and the combination of BP neural network model for diagnosis of wheat diseases.Consecutive samples of discrete data,applied to different matrix based on expert experience and heuristic calculation method RS reduction,and the reduction results as BP neural network input nodes.Experimental results show that this method not only optimizes the neural network topology,but also reduces the training time of neural networks,while greatly improving the learning speed and accuracy of fault diagnosis.