基于神经网络的大数据分析方法

Big Data Analysis Method Based on Neural Network

基于不同视角和情境特征的大数据定义诠释了大数据的5V特性,在扩展大数据价值空间与应用模式的同时催生了以"数据驱动+模型驱动"范式转变为代表的核心问题。为解决大数据分析核心问题,引入神经网络,采用性能优越的卷积神经网络设计对比实验,运用两个公开数据集对其进行训练,并在输出层分别使用L2-SVM和Softmax激活函数。在手写数字识别和彩色图像识别中,L2-SVM的识别错误率分别为0.87%和11.9%。实验结果表明,基于L2-SVM的神经网络大数据分析方法可以获得更高的识别精度。

The definition of big data based on different perspectives and situational features interprets the 5 V characteristics of big data. It not only expands the value space and application mode of big data,but also spawns the core problem represented by the transformation of"data driven + model driven"paradigm. In order to solve the core problem of big data analysis,the neural network is introduced,and the convolution neural network with better performance is used to design the contrast experiment. Two open datasets are used for training,and L2-SVM and Softmax activation function are used in the output layer. Among them,the recognition error rate of l2-svm is 0.87% and 11.9% respectively in handwritten digit recognition and color image recognition. The experimental results show that the neural network big data analysis method based on L2-SVM can achieve higher recognition accuracy.