Comparison of LSTM and GRU in Predicting the Number of Diabetic Patients
Eka Mala Sari Rochman, Miswanto Miswanto, Herry Suprajitno, Aeri Rachmad, Ratih Nindyasari, Fika Hastarita Rachman · 2022
Diabetes is one of the chronic diseases that many people have. This diabetes disease experienced a significant increase during the pandemic, which could cause numerous deaths. One way to help hospitals prevent too many diabetic patients is to predict the number of diabetic patients. We used the LSTM (Long Short-Term Memory) method to predict diabetic patients. The study was conducted using patient data from the Modopuro Health Center, Mojokerto Regency. The prediction process manually calculates the data, then looks for the correlation of the data according to the LSTM method, namely identifying the autocorrelation coefficients at two to three different time lags. The data observed is daily from January 2, 2021, to April 20, 2022, with as many as 345 data. From the calculation results, the RMSE value is 3.184, while the GRU produces an RMSE of 1.727. It concluded that the GRU could better predict the number of visits of diabetic patients in internal medicine polyclinics.