Workshop Goals

Predictive Analytics in healthcare poses several interesting technical challenges due to imbalance in data, high levels of accuracy expected and considerable missing information in the input, among others.  It is also a domain where real applications of analytics can create tremendous  impact. In particular, a Critical Care Unit (CCU) or an Intensive Care Unit (ICU) has the most critically ill patients in a hospital who are continuously monitored to check for disease progression and potential complications. CCUs are also data rich environments with a lot of heterogeneity in the available data – physiological measurements, radiology images and clinical notes are periodically recorded for each admitted patient. Although available in large volumes and velocity, the data is also fraught with uncertainties and noise. 

While a large number of CCUs exist, due to the 24x7 nature of the need for medical care, availability of resources – clinical  staff and monitoring equipment – are usually scarce. In addition, the cost of using a CCU has risen substantially amounting to a significant percentage of the overall hospital costs.  This motivates the growing need for predictive analytics solutions that can assist clinical staff in identifying high risk patients for prioritized and potentially life-saving care and also reduce healthcare costs.  The temporality and multimodality of data along with the need for robust and real time solutions have created new challenges in data mining and machine learning. 

The aim of this workshop is bring together researchers from the PAKDD community to address these challenges and design predictive analytics solutions for critical care.

The workshop will be collocated with the 20th Pacific Asia Conference on Knowledge Discovery and Data Mining (PAKDD) to be held between April 19 and 22, 2016. 
Venue: Auckland, New Zealand