基于外部知识辅助的人群健康数据预测方法

Population Health Data Prediction Method Based on External Knowledge Assistance

随着深度学习技术的发展和引入,现有人群健康数据预测方法的性能不断提高,但仍然受到数据质量问题的限制。为此,本文提出了一种基于外部知识辅助的人群健康数据预测方法。首先,该方法以与冠心病患病率相关性较强的高血压患病率数据和选区老年人口比例数据作为外部知识辅助填补冠心病患病率数据稀疏部分,对上述数据进行预处理后,构建CNN模型对高血压患病率数据和选区老年人口比例数据提取特征矩阵,并和随机噪声、部分完整的冠心病患病率数据作为CGAN模型的输入,以生成用来填补原冠心病患病率数据中稀疏部分的人工样本;然后,该方法将填补后的完整数据集通过ARIMA模型拟合得到模型特征,并输入GRU模型进行预测分析。实验结果表明,本文方法在MAE和RMSE上和KNN模型和RNN模型相差不多,但MPAE大大降低。

With the development of deep learning technologies,the performance of existing population health data prediction methods has been improved,but still suffers the limitation of data quality.In view of this,this paper proposes a population health data prediction method based on external knowledge assistance.In this method,firstly,the data of hypertension prevalence and elderly population proportion are utilized as external knowledge to fill the sparse part of coronary heart disease prevalence,due to their strong correlation,their feature matrixes are extracted via the CNN model and input into the CGAN model,with the complete coronary heart disease prevalence data and random noise part,to generate artificial samples;Further,the complete data set after filling is fitted by the ARIMA model to obtain the model features,and input into the GRU model for prediction analysis.The experiment results show that the proposed method has similar MAE and RMSE to RNN and KNN models,but less MPAE than them.