University Events

Description: Location: CCB-306
TITLE: MODELING AND PREDICTIVE ANALYTICS OF STREAMING HEALTHCARE DATA
Abstract: The amount of data in healthcare services is increasing with a high pace and is being generated every second. In the recent years, the healthcare systems around the world are experiencing fundamental transformation, as they move from a volume-based to a value-based healthcare delivery model. The data generated in healthcare services is mostly streaming data which can be in form of number, text or image. Medical researchers are using streaming data to speed up the decision making in hospital settings and improve healthcare outcomes for patients. The use of streaming big data has a potential to meet future market needs and trends in the healthcare organizations, it also gives an opportunity for physicians, epidemiologists, and health policy makers and analysts to make data-driven decisions that will ultimately improve patient care. By being able to quickly mine and analyze the huge amount of data and fulfil the growing demand is a challenging task This research aims to highlight modeling and performing predictive analysis on streaming healthcare data. One of such example is predicting freezing of gait (FoG) in Parkinson disease (PD). FoG is mostly frequent in the later stages of PD. Freezing of gait unable the patient to move the feet despite of his willingness to walk. This can lead the patient to fall and demotivate the patient to move. This disease can be treated with pharmacologic treatment but the effect of this drugs decrease, with the elapse in the duration of disease and also reduces mobility of the patient. Therefore, non-pharmacologic treatment is needed. In this research, we propose machine learning algorithms to detect the freezing in PD patients that can go along with non-pharmacologic treatments. Different data mining algorithms such as logistic regression, decision tree and random forest are applied to the data. We tested our algorithm on 10 patients suffering from PD. We performed user-dependent experiments on each patients and made a comparison on all the three algorithms based on sensitivity, specificity and misclassification error. This work would increase our basic understanding in streaming data in healthcare and will help towards assessing healthcare organization. In future, we will like to apply neural network on the same dataset and also like to focus our research in analyzing streaming text data in the healthcare services.

Advisor: Dr. Bharatendra Rai, Charlton College of Business
Committee members: Dr. Dan Braha and Dr. Gang Wang, Charlton College of Business;
Dr. Kristen Sethares, College of Nursing