search

alteryx / ta1-primitives

2 stars 3 forks source link

LL1_retail_sales_binary target data type issue impacts scoring #30

Open thehomebrewnerd opened 4 years ago

thehomebrewnerd commented 4 years ago

In the LL1_retail_sales_binary dataset, investigate why the target variable is not being read as a boolean when the datatype for the target variable is set to "boolean". With a "boolean" datatype, the target column is hashed and the scoring is incorrect.

When setting the target variable type to "categorical" the prediction pipeline works as expected.

thehomebrewnerd commented 4 years ago

This appears to be an issue with the d3m column parser common primitive as discussed here: https://gitlab.com/datadrivendiscovery/common-primitives/issues/124

One the d3m issue is resolved, we should verify the pipeline runs as expected with the target datatype set to "boolean".