Summary of Week 2
In this second week you have studied two different topics: Volunteered Geographic Information (VGI) and Self-organizing Maps (SOMs).
You have learned that more data leads to new types of analysis. Data mining techniques can be a solution to this problem. Some of these techniques belong to a group of analysis that we call Machine Learning.
Citizens can contribute to the production of geohealth data in massive amounts and at a fine spatio-temporal resolution. Volunteer data on geohealth can be used to feed complex scientific workflows based in machine learning algorithms. The results of these analysis may provide alternative perspectives for domain experts to design public health campaigns. Applying a spatio-temporal analysis requires three steps:
- Identify explanatory factors
- Model the geographic phenomena
- Visualize and interpret information
The two Dutch examples of tick monitoring showed us how volunteer information can be used to identify environmental conditions associated to tick bites and to create near real-time tick abundance maps.
Self-Organizing Maps are a so-called unsupervised learning algorithm. The term map is misleading as in fact it refers to a lattice of neurons and not a traditional type of map.
Applying Self-Organizing Maps is a twostep process:
- Training the SOM based on a training dataset
- Mapping data back onto the SOM
The example of Measles in Iceland showed us that SOMs allow us to compare different epidemics of Measles in order to compare the spatial-temporal diffusion pattens.
Hope to see you next week
We hope you enjoyed our course so far, and hope that you will stay with us in the next week. In week 3 we will focus on agent-based simulation and spatial statistics.
© University of Twente