1

In my data mining term project I need to conduct a work on geospatial data related to a big city to show possible urbanization fields in next years. I've used GRASS lately on other projects so I'm a bit familiar with GRASS.

After some research I've seen that R can be used for spatial statistics issues and I've read about forecast, raster, sp, spatstat, geospt etc libraries of R and it seems it can be helpful for my project. But on the other hand it has a very steep learning curve. Though I'm not sure that GRASS is whether sufficient or not at that point, learning process of R is seems scaring to me.

I'm coding for 4 years with C, Java, C#, Python etc but R's syntax, working mechanism is like an alien to me. What do you suggest for this problem. Do you think GRASS can be used solely or do i need to invest on R?

Thanks in advance.

EDIT: I have etm+ images produced at 3 different years(2010, 2011, 2012) for same location. I will use first 2 for training and last one for test. I will define parameters such as city centers, roads, urbanized places etc. to predict areas which are probable to be urbanized. There are some researches using Bayesian Belief Networks or Artificial Neural Networks to achieve this. I will use 2 or 3 parameters(roads, rivers) to estimate urbanization model of a city. Using vector data and raster data I think I can predict urbanization model more or less. So I need to evaluate raster cells against vector data to calculate their urbanization probability.

2
  • It will be difficult to give good, objective answers until you explain the nature of "conduct a work." Precisely what kinds of calculations do you anticipate conducting? What logical format will your "geospatial data" have? With what software will you need to interact, on which platforms? – whuber Jun 9 '13 at 16:02
  • Thank you, whuber. I've added some details in EDIT section. – GokcenG Jun 9 '13 at 16:23
1

In my opinion you should invest in R language. I saw some interesting solutions developed completely in R to study spatial evolution with cellular automata. Some of this studies are under developing in University of Eastern of Finland. Maybe you can try a contact to share expertise with them.

Some insights:

http://www.nrcresearchpress.com/doi/abs/10.1139/X07-073

http://www.sciencedirect.com/science/article/pii/S0378112706011042

I have not tried yet, but you have also some tools to mix qgis and R. Look this description:

ManageR is a QGIS plugin providing a simple and usefull interface to R statistical programming environment (http://www.r-project.org/). It is created by Carson J. Q. Farmer (http://www.ftools.ca/manageR) and is downloadable from this repository: http://www.ftools.ca/cfarmerQgisRepo.xml. To install it in QGIS is enough add such repository in QGIS Python Plugin Installer (Plugins → Fetch Python Plugins)

2
  • Thank you EBG. I've tried R, asked for my advisor, and postponed it to another time. It seems very complex but interesting too. I will look for more GIS-related tools for now. – GokcenG Jun 13 '13 at 8:00
  • @GokcenG, if the answer was useful and it helped you to make some decision accept it to closed your question. Best! – Gorgens Jul 15 '13 at 22:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.