Supervised Learning and Prediction of Spatial Epidemics

• Activity Number: 492
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Gyanendra Pokharel*+ and Rob Deardon
• Companies: University of Guelph and University of Guelph
• Keywords: Supervised Learning ; Spatial Epidemics ; Spatial Stratification

Abstract

Parameter estimation for mechanistic models of infectious disease spread can be highly computationally intensive. Nsoesie and coworkers introduced an approach for inference on infectious disease data based around the idea of supervised learning. Broadly, this method involved simulating epidemics from various infectious disease models, and then using a classifier built from the epidemic curve data to predict which model was most likely to have generated some other observed epidemic curves. We extend this work to the case where the underlying infectious disease model is spatial, and the nature of the spatial mechanism is unknown. A major goal of this study is to compare the use of global epidemic curves for building the classifier, with the use of sets of spatially stratified epidemic curves. Rectangular and circular stratification methods with various resolutions of stratification were used. Both stratification methods gave significantly better result up to a certain degree of stratification. The prediction error rate was optimized in different degree of stratification depending on the parameter sets. The circular stratification method outperformed the rectangular method.