Faculty Contact: John O' Donovan
When using spatial information, can we use data from neighboring areas to predict values for unknown areas? In which broader regions does this work well, and in which regions does this perform poorly? Throughout the world, there are many countries who engage in very little social, economic, and demographic data collection. While most countries will perform a census in various time intervals, some countries, whether due to political or religious conflict, are unable to collect any information on their citizens. When creating a global dataset for a given variable, the problem of imputing missing information can prove extremely challenging. Furthermore, information such as building characteristics can be critical in the event of a natural disaster, and if spatially explicit information of an area is not available, proper estimates must be used to inform emergency operations. The goal of this module is to explore just how well geo-imputation works in inferring characteristics of a given area. Changing the scale of this analysis, such as from city to county and national level, and measuring and visualizing the percent confidence in the predicted data at the varying scales, will be a primary goal for this project.
- TasteWeights: A Visual Interactive Hybrid Recommender System.
- Understanding Information Credibility on Twitter:
- Estimating the accuracy of geographical imputation:
- Missing in space: an evaluation of imputation methods for missing data in spatial analysis of risk factors for type II diabetes:
- The work of Ken Steif:
- Machine Learning for Geospatial Data:
- Organization of IPUMS data:
- Spring 2018: Su Burtner