Developing a Machine Learning Model for Real-Time Anomaly Detection in Patient Vital Signs Data

Patient vital sign data is crucial component of healthcare in recent years, which are the primary target for real-time monitoring, making a wide spectrum of research proposed machine learning models using vital signs for anomaly detection. This research concentrate a machine learning model to work as detectors in real-time for detecting the anomaly of vital signs of patient. This is a monitoring system which has been trained on Random Forest, Logistic Regression and Deep Neural Network to detect any abnormalities in major health matters (Heart rate, Blood pressure, Oxygen saturation etc.). The model filters streaming data sets, and incorporates feature engineering and data preprocessing techniques, all in a real-time application environment with low latency. Incorporating this model in healthcare systems will then facilitate such preventive measures by forewarning possible health risks. The performance of the model is accessed using accuracy, precision and recall which give high promise in improving human patient safety by continuous monitoring. This research adds towards the growing body of works in health informatics and predictive analytics.