利用支持向量机和人工神经网络填补缺失数据

The Use of Support Vector Machines and Artificial Neural Networks to Fill Missing Data

摘要: 本文从R内置数据集iris中按需要选取样本数据建立学习样本，模拟生物样本属性值缺失和类别缺失两种缺失数据的情况，以MATLAB为工具，利用支持向量机和人工神经网络对缺失值进行填补。对于生物样本数据中存在属性值缺失的情况，可以分别采用支持向量机和人工神经网络进行回归填补，并对BP神经网络和RBF神经网络的适用性进行了对比；对于生物样本数据中存在样本类别缺失的情况，采用支持向量机分类填补。结果显示，用神经网络预测填补缺失的属性值时，RBF网络对隐层神经元数目选取的自适应性使之比BP网络更为稳定；相比人工神经网络，支持向量机对有限的样本更为适用，并且不依赖设计者经验，泛化能力强。

Abstract: In this paper, we use the built-in data set iris to select the sample data according to the need to es-tablish the learning samples, simulate the missing values of the biological sample attributes and the missing two types of missing data. Using MATLAB as the tool, we use the support vector machine and the artificial neural network to carry out the missing values to fill. For the case of missing attribute values in the biological sample data, the support vector machine and the artificial neural network can be used for the regression filling, and the applicability of the BP neural network and the RBF neural network can be compared. For the sample of the biological sample, missing situation, using support vector machine classification to fill. The results show that the adaptability of the number of hidden neurons in the RBF network is more stable than that of the BP network when the neural network is used to predict the missing attribute values. Compared with the artificial neural network, the support vector machine for the application, do not rely on the designer experience and has generalization ability.