The Data Mining Forum                             open-source data mining software data science journal data mining conferences high utility mining book
This forum is about data mining, data science and big data: algorithms, source code, datasets, implementations, optimizations, etc. You are welcome to post call for papers, data mining job ads, link to source code of data mining algorithms or anything else related to data mining. The forum is hosted by P. Fournier-Viger. No registration is required to use this forum!.  
Problem: which parts of a raster map determine occurrence of an event?
Posted by: Duberdico
Date: July 20, 2017 09:05AM

Dear All,

I am working on a problem where I would like to assess if variation in particular regions of a map determine the occurrence of an event.

I have raster maps (aprox 200x400 cells) of the distribution of a parameter in space. I have one of these maps every year for a range of years (about 20). In some years a particular event occurs where in other it doesn't. I hypothesize that the occurrence of this event may be caused by variation in a particular region of the map , e.g. in some years a particular region of the map may take higher values leading to the occurrence of the event.

I am currently drawing blanks thinking of a method that could be used to tackle this problem. I would greatly appreciate if someone could suggest an approach.

Thank you in advance for any comments and suggestions.

Options: ReplyQuote
Re: Problem: which parts of a raster map determine occurrence of an event?
Posted by: Phil
Date: July 20, 2017 04:21PM

I think it can be seen as a classification problem, where given a map, the goal is to predict whether it belongs to class 1 (event) or class 2 (no event).

For a classification problem, you could use some classifier such as neural networks. However, having only 20 maps is a very small training set of data for training a neural networks. But maybe it could work. Or depending on the data, maybe it is possible to use a more simple type of models.

I think that a key problem is how to define the features for training a model. You could use every pixel as an input, or you could define higher level features based on these pixels. It depends on the application.

This is just some ideas.

Options: ReplyQuote


Your Name: 
Your Email: 
Subject: 
Spam prevention:
Please, enter the code that you see below in the input field. This is for blocking bots that try to post this form automatically.
  ******         **  ********  **        ******** 
 **    **        **  **    **  **        **       
 **              **      **    **        **       
 **              **     **     **        ******   
 **        **    **    **      **        **       
 **    **  **    **    **      **        **       
  ******    ******     **      ********  **       
This forum is powered by Phorum and provided by P. Fournier-Viger (© 2012).
Terms of use.