ML Pipeline Node Regression missing predict_write

Neo4j Graph Platform Graph Algorithms/Graph Data Science
,
1

I'm using the Neo4j Graph Data Science Client (Python) to run a ML Node Regression pipeline. The resulting model (graphdatascience.model.node_regression_model.NRModel) contains methods for predict_stream and predict_mutate however it doesn't contain a predict_write method.

Is this by design?

NRModel is in file graphdatascience/model/node_regression_model.py

p = gds.alpha.pipeline.nodeRegression.create("NR_Pipeline")
p.train(...)

The resulting model from the train method supports predict stream and mutate but there is no predict_write. Is it missing by design?

2

Hi @RichardBonnett ,

There is no write mode currently; only stream and mutate are supported:

You can look into using the following python command for writing back to the database.

which is the front-end of Writing node properties and labels - Neo4j Graph Data Science

Best regards,
Ioannis.

3

Thank you Ioannis,

Do you know if this is a permanent 'no write mode', maybe due to some limitations. Or might it be added later and I should look out for it.

Thank you.

4

The pipelines are in alpha tier and thus are not feature-complete yet.

I unfortunately do not know when or if write capabilities will be added. Pipelines are not a big focus for us right now :confused:

For the moment, I would suggest you workaround it via the gds.nodeProperties.write even if it is slightly inconvenient...

Best,
Ioannis.