CAJ | 학술논문

为预测糖尿病患者未来一段时间内的血糖值，提出一种融合差分自回归移动平均（A RIM A ）和径向基神经网络（RBFNN）的组合预测模型。利用ARIMA对患者未来血糖值进行预测，采用RBFNN修正预测结果。针对外部因素对血糖的影响，提出一种发现和处理算法。采用医院数据对模型和算法进行验证，验证结果表明，相比A RIM A预测，提出的模型具有更高的预测精度；提出的算法能准确处理外部因素对血糖的影响，保障预测精度。
위예측당뇨병환자미래일단시간내적혈당치，제출일충융합차분자회귀이동평균（A RIM A ）화경향기신경망락（RBFNN）적조합예측모형。이용ARIMA대환자미래혈당치진행예측，채용RBFNN수정예측결과。침대외부인소대혈당적영향，제출일충발현화처리산법。채용의원수거대모형화산법진행험증，험증결과표명，상비A RIM A예측，제출적모형구유경고적예측정도；제출적산법능준학처리외부인소대혈당적영향，보장예측정도。
To predict future glucose values of diabetics ,a hybrid autoregressive integrated moving average‐radial basis function neural network (ARIMA‐RBFNN) model was proposed for short‐time glucose prediction .ARIMA was used to forecast future glucose values ,and then RBF neural network was used to correct the prediction .At the same time ,an algorithm was proposed to deal with the abnormal changes in blood sugar that influenced by external factors .The results of numerical example based on the data from hospital show that the proposed hybrid ARIMA‐RBFNN model outperforms the ARIMA model and the proposed algorithm can handle the influences caused by external factors .As a result ,it can ensure the forecast accuracy .