基于PCA融合神经网络的移动设备威胁研究

Study of mobile device risk base on PCA and neural network

为了提高移动设备存在威胁风险分析的准确率和抗干扰能力,进而降低投诉率,提出一种基于PCA融合BP神经网络风险分析模型。通过核主成分分析将设备威胁类型从10个特征量降低到3个主特征量,提取了原始数据的主信息,并以降维后的特征量作为BP神经网络的输入特征量,建立设备威胁风险分析模型。最后通过实验对比了多种模型算法,结果表明采用PCA结合BP神经网络的风险分析模型具有更好的风险识别准确率。

To improve the accuracy and anti-interference ability for risk analysis of mobile device, and reduce the complaint rate,this paper proposed a model of risk analysis base on PCA and BP neural network.The feature vectors of device risk types are reduces from 10 to 3 and main information of raw data can be got by PCA. And then built a device risk analysis model using BP neural network which the input data is the low dimension feature vectors got from PCA. Experiments compared performances of our model and some other models, the results show the model combine PCA and BP neural network can identify risks accurately.