基于Word2vec和卷积神经网络特征提取的双高疾病预测

DOUBLE-HIGH DISEASE PREDICTION BASED ON WORD2VEC AND CONVOLUTIONAL NEURAL NETWORK FEATURE EXTRACTION METHOD

在高血压和高血脂疾病的预测研究中,针对体检数据中文本型数据特征提取问题,提出利用Word2vec和卷积神经网络相结合的方法(WV-CNN)对数据中的文本特征进行特征提取,建立预测模型。利用Doc2vec方法进行特征提取的对比实验,结果证明该预测方法的特征提取能力在不同输入数据数量级和不同预测方面都有很好的表现,对双高疾病识别和预测效果较好。

In the predictive study of hypertension and hyperlipidemia,aiming at the feature extraction of textual data in physical examination data,a method combining Word2vec and convolutional neural network(WV-CNN)is proposed to extract features of text features in data,and a predictive model is established.The comparison experiment of feature extraction using Doc2vec method shows that the feature extraction ability of WV-CNN method has a good performance in different number level of input data and different methods of prediction,and the double high disease identification and prediction effect is better.