基于深度神经网络的医院门诊量预测

A Study on Outpatient Prediction Based on Deep Neural Network

为了有效地预测医院门诊量,充分考虑历史门诊量数据和工作日天数之间的关系,提出一种应用深度神经网络预测方法,深度神经网络模型由RBM层和预测层组成,采用无监督学习算法预训练网络参数,引入残差结构使输入信息跨层传输,利用反向学习算法微调网络参数,进而获取优化后的深度神经网络预测模型。实验结果表明,深度神经网络模型经过2层RBM训练之后,即可从原始样本中提取代表性较强的数据特征,所提方法在小样本数据下可以获得较好的预测精度,能够为医疗业务规划提供理论参考。

To effectively forecast the outpatient in a hospital,by fully considering the relationship between historical outpatient volume data and working days,a deep neural network prediction method is proposed.The model consists of RBM layers and a prediction layer.An unsupervised learning algorithm is used to pre-train network parameters.A residual structure is introduced to enable input information to be transmitted across layers and use the reverse learning algorithm to fine-tune the network parameters,and then obtain the optimized deep neural network prediction model.Experimental results show that the deep neural network model can extract representative data features from the original samples after two layers of RBM training.Compared with the traditional prediction model,the proposed method can obtain better prediction accuracy under small sample data,which provides theoretical reference for medical business planning.