Medical Data Mining on Abdominal Aortic Aneurysm Patients' Dataset:
Discovering the Relationships between the Medical Records and the Length of Stay at the hospital (LOS)
Pooya Nikbakht - Summer 2014

Medical Data Mining: Discovering the Relationships between Abdominal Aortic Aneurysm Patients' Medical Records and  Length of Stay at the hospital

Abstract: The Abdominal Aortic Aneurysm (AAA) patients' dataset (including 6000 records produced by Vascular Governance North West of England from years 1997 to 2012) was analyzed by using data mining techniques to deduct valid relationships between the patient's medical records and the length of hospitalization. At first, considering the sensitiveness of medical data and the existence of lots of noise, outliers, and missing values, an accurate preprocessing stage was done to generate a clean and valid dataset. Then, different model learning methods such as Decision Tree (C4.5), Rule Induction, KNN, and Naïve Bayes were applied to the dataset. Consequently, their accuracy and validations were assessed and due to the precise pre-processing, some high accuracy models (between 70 to 84 percent) resulted. And finally, by merging the decision tree and rule induction results, some meaningful rules were inferred and proposed which can help medical managers better manage the resources.


This work was done as my undergraduate's final project under the supervision of Professor Mo Saraee and graded 20/20 (referee: Dr. Abdolreza Mirzaei).