基于MATLAB的手写体数字识别算法的现与分析-毕业论文.docxVIP

基于MATLAB的手写体数字识别算法的实现与分析摘 要手写体数字识别是利用计算机自动辨认手写体阿拉伯数字的一种技术，是光学字符识别技术的一个分支。手写体数字识别在邮政编码、财务报表、银行票据、各种凭证以及调查表格的识别等等方面有着重要应用，由于数字识别经常涉及财会、金融领域，其严格性更是不言而喻的。所以，对识别系统的可靠性和识别率要求很高，构成了手写体数字识别面临的主要困难，大批量数据处理对系统速度又有相当高的要求。本文基于MNIST数据集，通过Matlab平台，对决策树算法、SVM算法和人工神经网络（ANN）算法进行实现，并对分类算法的准确率进行评估。实验结果表明，人工神经网络（ANN）的准确率最高，为99.69%，SVM算法次之，准确率为94.53%，决策树算法的准确率为83.53%。三种分类算法中，决策树算法的速度最快，SVM算法的速度最慢。另外，针对每一种分类算法在MNIST数据集上的实验结果，本文还得出以下结论：第一，MNIST数据集的归一化与否对决策树的分类效果几乎没有影响；对SVM的分类效果影响较大，未归一化时的准确率为11.35%，归一化之后的准确率为94.53%；对人工神经网络的分类效果影响较小，未归一化时的准确率为82.11%，归一化之后的准确率为99.69%。这说明三种分类算法对数据的不平衡分布的敏感程度各不相同。第二，对于SVM分类算法，当训练数据集的样本容量小于60000（MNIST训练数据集的最大样本容量）时，该算法对测试数据集分类预测的准确率随样本容量的增大而增大。第三，针对人工神经网络，数据类标签的表示形式对分类预测的准确率的影响较大。使用10位数据表示类标签是的准确率为99.69%，远远高于使用1位数据表示类标签时的准确率60.24%。关 键 词：手写体数字识别；决策树算法；SVM算法；人工神经网络算法 ABSTRACTHandwritten numeral recognition is a technique that uses computer to recognize handwritten Arabic numerals automatically and is a branch of optical character recognition technology. Handwritten numeral recognition has important applications in postal codes, financial statements, bank notes, various kinds of vouchers and the identification of survey forms. Since digital identification often involves accounting and finance, its strictness is self-evident. The demand for identification system of the reliability and recognition rate is very high, constituting a handwritten digital identification facing major difficulties, high-volume data processing on the system speed and a very high demand.In this paper, we use Matlab to implement decision tree algorithm, SVM algorithm and artificial neural network (ANN) algorithm based on MNIST dataset, and the accuracy of classification algorithms is calculated by using the real data tag. Experimental results show that the artificial neural network (ANN) the highest accuracy rate for 99.69%, SVM algorithm, followed by 94.53 percent accuracy rate, decision tree algorithm accuracy is 83.53%. In terms of speed, decision tree algorithm is the fastest, SVM algori