0

Value of each variable in MLP neural network

ghilman

April 23, 2013 17:43

None

Follow

I tried to run Land change modeler, i use MLP Neural Netwk.

the explanatory power of a variable can be test by Cramer’s V. but in "help" explaned that A high Cramer’s V indicates that the potential explanatory value of the variable is good, but does not guarantee a strong performance

After running the model with MLP Neural Netwk.

How we know the real value of the influence of each variable on land cover changes?

eg the comparison between the road and the river to change non-urban to urban...

Thanks

• Facebook
• Twitter
• LinkedIn

2 comments

• 0
  
  Clark Labs April 24, 2013 19:46

  Ghilman,

  If you have not done so already, you should upgrade to Selva 17.02. If you have Selva 17.00, this is a free upgrade. Once you upgrade, the issue of variable importance with MLP has been solved. From our website:

  Perhaps the most substantial revision of this release is to Land Change Modeler (LCM) and to the creation of the transition potential model using the multi-layer perceptron (MLP) neural network option. We have resolved the major issue of exposing the explanatory power of the independent variables and opening the black-box associated with this routine. The user now has feedback on both the general model and the independent variables. General model information now includes a skill measure to accompany the overall accuracy level of the model. The skill measure is based on the analysis of the validation data. The most significant addition is a sensitivity analysis of the independent variables used in the model. This analysis evaluates the quality of the independent variables and those that could be excluded from the model.

  This new facility should address your concerns.

  James Toledano

  Clark Labs

• 0
  
  ghilman April 25, 2013 04:36

  Thanks for attention...

  Yes i use selva 17.00.

  in your website to Upgrade to Selva 17.02, required selva 17.01.

  what i should do ?