search

getml / getml-community

Fast, high-quality forecasts on relational and multivariate time-series data powered by new feature learning algorithms and automated ML.
https://getml.com
Other
93 stars 9 forks source link

[Question]Add time constraits when a middle table has no timestamp #11

Open DanielMao1 opened 2 weeks ago

DanielMao1 commented 2 weeks ago

In snowflake schema, How to set constraits if an intermediate table has no timestamp. Hello, I am a getML user. I am trying to switch from featuretools to your tools. I appreciate this great job you did. But I encounter a problem: suppose in some of datasets which is snowflake schema, T0 is target table, and there is a join path: T0->T1<-T2. If T0, T1, T2 all have timestamp, that's OK, we can add constraints on join like

schema.join(
    T1,
    on="FK", 
    time_stamps=("date1",  "date2"),
)

A more concrete example is in this schema, customers and calls have timestamp, but complaints does not. How can I add time constraints on customers and calls ? snowflake_schema

But if T0 and T2 has timestamp, but T1 does not, I can not set timeconstraits althouth I would like to because T1 has no timestamp. It works fine on featuretools but does not work on getML. Do you plan to support such thing or are there any workwround?

Jogala commented 2 weeks ago

Hi @DanielMao1, thanks for your question.

As I quick fix, I would suggest to add the customer_id to the calls table and then directly joining the calls table onto the customers table.

Could you please share the feature tools code?

Is that a public source dataset?