Outliers

devonkyle

Is there a way within GeneXpro to preprocess the data for removing Outliers ? The Data tab charts (scatter plots etc) are very useful in finding these issues, but I'm not seeing a method to remove or modify them . I'm assuming, like most modeling programs, outliers can have a strong impact on the modeling processes within GXP..

Thank you

Devon Kyle

GepsoftTeam

Hello,

Indeed outliers can impact negatively the quality of your models. After you identify the outliers you can remove those rows using the menu Data -> Delete Records. Here you can delete one or more records:

To delete one record just enter its Id, to delete several enter the Ids separated by spaces.

devonkyle

and using the Highlight Records / POINT option helps in detecting those. I'm seem to becoming an expert at answering my own questions on this application ...:>

Loving what I think is a new feature in 5.0 Preview . Rows on the RESULTS tab are highlighted that are wrong. Can't remember if the previous version had that or not, but it's very useful. Very clever marketing by the way, getting us hooked on new features :>

Here is where I'm stuck. Choosing the Fitness Function.
I'm a little lost on how to experiment with these. I realize that modeling is not an exact science and a lot of discover involves trial / error, testing etc. What would you recommend as a good resource to get a better idea of the fitness functions operators without having to get a degree in stats (something I now wish I had done).Do certain functions lend themselves better to certain types of data better than others ? I know this is a very broad question that can't be easily answered in a forum, but if you have any tools to get me in the right direction, I'm willing to do the homework..

Thanks

Devon

Dr. Candida Ferreira

Hi Devon,

Trial and error is fine, especially if you can see straightaway the result of what you’re trying, which in GeneXproTools you can just by looking at the charts during evolution. I think this will be faster than getting a degree in statistics :) But anyway, you can take a look at the section on Fitness Functions in an article I wrote for Wikipedia a while back:

http://en.wikipedia.org/wiki/Gene_expression_programming#Fitness_functions_and_the_selection_environment

Candida

devonkyle

Thanks for the share Candida! This looks like something I could actually (sorta) understand :>

Howdy, Stranger!

Categories