Community Spring Cleaning week is here! Join your fellow Maveryx in digging through your old posts and marking comments on them as solved. Learn more here!

Alteryx Designer Desktop Discussions

Find answers, ask questions, and share expertise about Alteryx Designer Desktop and Intelligence Suite.
SOLVED

Linear Regression: several error messages

cdahl001
7 - Meteor

Good morning! 

 

I'm trying to predict the likelihood of product being rejected upon receiving based on weather conditions of the supplier location when filling orders. I'm experienced with Alteryx, but not with the predictive tools. I dabble in several programming languages, but not R. These errors make no sense to me. Can someone point me in the right direction for error resolution? 

 

Linear rejection errors.JPG

 

5 REPLIES 5
RobertOdera
13 - Pulsar

Hi, @cdahl001 

 

If you hadn't already done so, kindly wash the data and re-attach a sample file as a packaged workflow (so we can attempt to help you).

The current yxmd attached means that we can't run the workflows we can't path to the file location on your desktop/ laptop etc.,

 

A "subscript out of bounds" usually means that Alteryx is trying to access an object that isn't there.

Make sure that all the fields that you need as inputs are flowing into the process/ tool.

 

Maybe you have the wrong field type so the field dropped because (for example) the algorithm needs to ingest a string whereas you are providing a different field type. I will offer that you add a Select Tool before input node into Regression Tool and use it to inspect the fields and field types to see if something needs tweaking or added or dropped.

 

The best case, however, is to attach a sample so that we can work it for you.

cdahl001
7 - Meteor

@RobertOdera Oh! Sorry about that. Here you go: 

 

 

RobertOdera
13 - Pulsar

Hi, @cdahl001 

 

It looks like all the predictor fields have a significant %missing values that are flowing through as Nulls.

I will offer that the Nulls are throwing the errors.

Kindly consider imputing the values.

 

It might also be that your data set lacks richness/ density (some of the fields have a very small number of value counts).

Kindly consider combining some values/ aggregations if and where it makes sense.

 

Finally (not sure if this is a factor), your field names have spaces.

In my experience, I've seen algorithms struggle with that.

Kindly consider using the Dynamic Rename tool by configuring it per the snippet below.

 

RNO2_0-1587584406125.png

 

 

 

cdahl001
7 - Meteor

That makes sense! I did not look for those things and will remedy that now. 

 

Any idea how many rows of data are needed to get a good model? I only have so much data to work with, but can make fake data to feed it. 

RobertOdera
13 - Pulsar

Hi, @cdahl001 

 

Why don't you DM me and let's take it from the top (let's build the data set from scratch from the inputs that you have) together?

We can zoom or Teams.

Once we have a way forward (or not), you can post learnings on here to benefit the Community.

How much data you have is a critical predictive approach criterion.

The outcome you're after will determine if imputation (what measure of central tendency to use) or omission is the way to go.

Also, it doesn't have to be perfect (good enough will do!).

 

Labels