注意力机制结合卷积神经网络的医院财务风险预警方法研究

Research on early warning method of hospital financial risk based on attention mechanism and convolutional neural network

面对医院可持续性经营能力有效管理和预警问题,文章构建了基于注意力机制和卷积神经网络的医院财务风险预警模型,运用狼群算法进行模型优化,来减少模型的输出误差。对初始数据进行清洗、特征选择、归一化的预处理,以及进行样本生成,在此基础上,通过该模型对医院财务数据进行分析,以此来为决策者提供相关参考数据。实验结果显示,相较于卷积神经网络、KNN算法等算法,文章采用方法的性能更佳,在50次的迭代次数下,其预测准确率最高为93.47%。训练数据量影响着文章采用模型准确率,当u值较大时,准确率在0.85附近波动。在医院财务风险预警中,文章采用方法训练集的拟合优度为92.83%,高于其他算法。文章获得的风险预警数据对决策者有一定的参考价值。

Facing the problem of effective management and early warning of hospital sustainable operation capability,this paper constructs a hospital financial risk early warning model based on attention mechanism and convolutional neural network,and uses wolf colony algorithm to optimize the model to reduce the output error of the model.The initial data are preprocessed by cleaning,feature selection,normalization,and sample generation.On this basis,the hospital financial data are analyzed through the model to provide relevant reference data for decision-makers.The experimental results show that,compared with convolutional neural network,KNN algorithm and other algorithms,the performance of the method adopted in this paper is better,and its prediction accuracy is 93.47%at 50 iterations.The amount of training data affects the accuracy of the model adopted by the article.When the u value is large,the accuracy fluctuates around 0.85.In the early warning of hospital financial risk,the goodness of fit of the method training set used in this paper is 92.83%,which is higher than other algorithms.The method can provide effective reference data for decision-makers.