Infusion pumps in hospitals are used to transfer drugs intravenously to the patient. These devices deliver drugs by Preprogramming the infusion rate. Hence, a constant amount of drug is delivered to the patient without continuously monitoring their vital changes. Various medication errors have been reported due to this preprogrammed infusion, especially for blood pressure (BP) control using inotropic drugs. The BP variations of patients are highly nonlinear and time variant. Due to this complexity, there is a lack of automated controllers to adjust the infusion rate with changes in BP. We developed an automated closed-loop approach utilizing a long short-term memory (LSTM) network to predict infusion rates based on the changes in BP. For an initial study, we collected data from ICU patients who were supported with noradrenaline. We forecast the infusion rate and compared the performance of the LSTM model by using various windows of prior input data. Statistical parameters such as mean absolute percentage error (MAPE), coefficient of determination ( \(R^2\) score), mean absolute error (MAE), standard deviation (SD) of error, and root mean square error (RMSE) show that the model performance improved using 15 min of prior data. The results demonstrate that a shorter time window of 15 min captures the recent BP variations, allowing the model to predict noradrenaline rate more accurately.