Keywords: Spatial Data Mining, Big Data, Co-Location Mining, Data Analytics, Data Integration, Health Informatics.
The Department of Computing Science and the Department of Paediatrics at the University of Alberta, Canada, anticipate hiring a postdoctoral fellow for one year and renewable for a second year. The position includes a full-time salary and benefits, and is funded by the Domino Project.
Location: University of Alberta, Department of Computing Science and Department of Paediatrics
Project: DOMINO (Data Mining and Newborn Outcome Project)
Title: Spatial data mining exploring co-location of adverse birth outcomes and environmental variables.
Abstract: There exists a multitude of available publicly funded databases related to chemical releases by industry, related to neonatal and perinatal health, as well as various statistics and weather and environmental data. We would like to recruit a postdoctoral researcher for one year (or longer) to work with a multidisciplinary team and address problems specific to the integration of these heterogeneous data sources, model the data and apply data mining approaches to discover co-location patterns and other useful hidden patterns in the data.
Context: This interdisciplinary research project investigates chemical and socioeconomic environmental influences on maternal/infant birth outcomes (low birth weight, preterm births, stillbirths and perinatal mortality). The team comprises researchers in computer science, medicine, public health, environmental studies, earth and atmospheric science, etc. Some preliminary work has already been conducted by graduate students from various disciplines on some available data devising specific algorithms to extract patterns. The postdoctoral fellow candidate would consolidate between these works, innovate beyond the preliminary work, assess off the shelf algorithms and tools for patterns discovery on the available spatial data and bridge between the team members.
Description: Spatial data mining is the process of discovering potentially useful patterns from large spatial datasets. Due to the complexity of spatial data, spatial data mining can be more difficult than extracting the same patterns from conventional data, owing to the presence of spatial relationships and autocorrelations. Typical spatial pattern mining tasks include spatial association rule mining, co-location mining, spatial outlier detection, location prediction, etc. Many algorithms exist but each spatial data application has its own idiosyncrasy requiring adaptation or the creation of new algorithms particularly due to the data integration and data modeling dictated by the application. The candidate will be in charge of assessing existing algorithms and discovering new efficient spatial data mining techniques for extracting spatial patterns, and designing pattern visualization tools to help practitioners assess and validate the discovered patterns.
Supervision: The project will be formally jointly supervised by Dr. Osmar Zaiane at the Department of Computing Science and Dr. Alvaro Osornio Vargas Department of Pediatrics
Starting date: As soon as possible
Conditions of employment: