While scoring one of my prediction models I encountered the following error:
sh: sns-cmd: command not found
Any pointers on how to resolve this?
Can you give more details? Can you give a screenshot of the log and your R snippet?
I've added you to my insights org. You can view the log online here. The R script is an adaptation of the sample script pointing to my catalog instead of the default RetailDatasets catalog.
Thanks for adding. Can you provide me your R script for model and score?
Answer by amalavalli@apigee.com · Apr 14, 2016 at 10:07 PM
Basically, there is no correlation found in the model. modelObject$stream(topN = -1) should tell you that if there are like 2 or 3 lines only. My guess is that it has to do with dimension value in setProfile method. It looks like you might have something like setProfile(<OTHER PARAMS>, dimensions = list(c("AgeGroup","Gender","IncomeGroup","MaritalStatus","State","HouseholdSize")))
change it to
setProfile(<OTHER_PARAMS>, dimensions = list("AgeGroup","Gender","IncomeGroup","MaritalStatus","State","HouseholdSize", c("AgeGroup","Gender"),c("AgeGroup","IncomeGroup"), c("AgeGroup","MaritalStatus"), c("AgeGroup","State"), c("AgeGroup","HouseholdSize"), c("Gender","IncomeGroup"), c("Gender","MaritalStatus"),c("Gender","State"),c("Gender","HouseholdSize"), c("IncomeGroup","MaritalStatus"),c("IncomeGroup","State"),c("IncomeGroup","HouseholdSize"), c("MaritalStatus","State"), c("MaritalStatus","HouseholdSize"), c("State","HouseholdSize")))
Basically, in your script you are telling the model to use:
AgeGroup and Gender and IncomeGroup and MaritalStatus and State and HouseholdSize as *one combination* and try to find correlation. This might not have enough support since this could be a too narrow of a marker.
In the change that i have suggested, we are telling the model to use:
AgeGroup or Gender or IncomeGroup, ....
OR
AgeGroup and Gender
OR
AgeGroup and IncomeGroup
OR
etc
This way we try to find correlations for each combination. So the bottom line is that, usually with experiments we have done so far, 2 or max of 3 dimensions as one combination works well. But ultimately it depends on the data.
This might be little confusing at first. And the tutorial does not talk about this actually.
Thanks for the detailed answer and suggesting the changes in the script. I'll rerun the script and let you know the outcome. Could you meanwhile convert your above comment to an answer so that I can accept it?
I converted the comment to an answer.