Specifying Custom Missing Value Representations

There may be cases where your data contains missing values, but they are represented in a nonstandard way (other than null or NaN). As an example, suppose an upstream system uses "-1.0" or "-999" in place of null for a particular Float column. Fiddler offers a method to specify your own missing value representations for each column when defining your model schema. See below for an example.

---

You can modify your fdl.ModelSchema object just before onboarding your model to include details about which values should be replaced with nulls when publishing data to Fiddler.

model.schema['my_column'].replace_with_nulls = [
  '-1.0',
  '-999'
]