Lift Chart Tool - True response rate settings

[FredrikHogstrom]

Hi all,

I have a question about the Lift chart example workflow (Predictive Analytics -> 10_Lift_Chart.yxmd). 

The input data set seems to have a 30 percent response rate (Default = "yes"), but in the lift chart tool the true response rate is set to 0.05. Can anyone explain where the 0.05 comes from?

Best regards,

Fredrik 

[fadib · Accepted solution]

Hi Fredrik,

The data set was oversampled prior to being loaded into Alteryx (which is not clear from the sample, the orignial data had 5% YES responses) and therfore the level was set to the original sample rate, not the data as available.

[akaur]

I'm having some similar questions on this.  I'm not very clear on this.  I should have true response rate of what the response rate is for my campaign?

Also, I am getting an error "missing incoming connection" on my Lift Chart.  I oversampled and created a sample and have the lift chart but don't know what this error is.

[akaur]

I'm having some similar questions on this.  I'm not very clear on this.  I should have true response rate of what the response rate is for my campaign?

Also, I am getting an error "missing incoming connection" on my Lift Chart.  I oversampled and created a sample and have the lift chart but don't know what this error is.

Need some help, please.

[fadib]

Hi @akaur

You are correct, for the Lift-Chart you need to input the True Response Rate (of the target value ex. for "yes" if you're trying to predict the "yes") before any Oversampling. You can check the true response rate using the Frequency Table tool from the Data Investigation palette before any oversampling you're doing. In the above workflow, the data was oversampled before being brought into Alteryx and therefore the workflow already has oversampled data at the input (which might be a bit confusing).

on the second question, we'll need to look at the workflow. If you can, upload it to this thread; or if you prefer, email it to support@alteryx.com and one of our customer support engineers will be in touch. 

[FredrikHogstrom]

Great, thanks for the reply.

Fredrik

[Inactive User]

Hi @fadib,

I just saw you answer. So if my original data set have Response Rate 10% with millions of rows. So I create a sample data which has half millions of rows and keep the response rate at 20%. 

Then I use this data set to training model with sample tool with 70% training and 30% . 

In the lift chart tool, 

if I use the original data set as input (Response Rate 10% with millions of rows), what  response rate should I use?

if I use the 70% training data set as input(Coming from half millions of rows with response rate at 20%), what  response rate should I use?

Thank you so much.