Multivariate Time Series Prediction of Pediatric ICU data using Deep Learning

Abstract

The pediatric cardiac intensive care unit (ICU) is a specialized section for children with heart diseases. The patients admitted to the ICU are in a very critical condition. The data for each day were collected hourly basis. So, the time-series prediction might be beneficial for the physicians for the medication process of the patients whose lives are in danger. This paper proposes a multivariate time series prediction where multiple features with respect to timestamps are to be predicted using the deep learning methods in order to assist doctors in decision making in the tensed moment. Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) these methods are applied for the time series prediction. The comparative analysis among the RNN and LSTM prediction model is also highlighted in this paper. Doctors' advice is also taken to justify the result.