A study of kNN using ICU multivariate time series data

<p>The purpose of this research is to study the performance of kNN (k Nearest Neighbor) classification approach to determine patients’ mortality rate using ICU (Intensive Care Unit) medical records. The ICU data contains medical records collected during the patients’ first 48 hours stay at the ICU. The challenge of this research is the processing of ICU multivariate and high dimensional time-series data collected at irregular time periods. To handle the ICU irregular multivariate time-series three different methods were developed: Capture Statistics, Detect Changes, and Aggregate Segments. We examine the effectiveness of each method on kNN classification. In addition, this paper addresses imbalanced class distributions and their effect on kNN performance.</p>

Name thesis:degree_name

Master of Science (MS) in Computer Science

Level thesis:degree_level

Thesis: EWU Only

Discipline thesis:degree_discipline

Computer Science

Year

2014

Author dc:creator

• Djulovic, Admir

Contributors dc:contributor

• Dr. Dan Li
• Dr. Carol Taylor
• Dr. Robin O'Quinn

• Computer Sciences

Statement dc:rights

• Access perpetually restricted to EWU users with an active EWU NetID

Repository record dc:identifier

https://dc.ewu.edu/theses/262

OAI identifier oai:identifier

oai:dc.ewu.edu:theses-1261