question

NadineBenharrath-4116 avatar image
0 Votes"
NadineBenharrath-4116 asked YutongTie-MSFT commented

how can i retrain the model after a period of time

Hello everyone, i'm using lambda architecture to build a fraud detection project , i build my model using machine learning in databricks , after saving the model , i load the model in the speed layer to predict the incoming data, i want to know how can i retrain this model using new incoming data from eventhub ??
does the retrain should be in the batch layer ?
thanks for helping

azure-machine-learning
· 1
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

@NadineBenharrath-4116 Thanks for the question. Can you please add more details about the model that you are trying.

0 Votes 0 ·
ramr-msft avatar image
0 Votes"
ramr-msft answered ramr-msft edited

@NadineBenharrath-4116 Thanks, Please follow the document to retrain the model using the new data.

Currently Data Drift Monitor (Data Drift->EventGrid->LogicApp->Trigger Pipeline) allows to trigger a pipeline when dataset drift has been detected.

Public Repo of the architecture and code: https://github.com/Microsoft/MLOps


5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

NadineBenharrath-4116 avatar image
0 Votes"
NadineBenharrath-4116 answered YutongTie-MSFT commented

@ramr-msft the problem that i trained my model using databricks and i tracked it with mlflow , i'm not using azure ML because my dataset is imbalanced ?

· 1
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

@NadineBenharrath-4116 Could you please check on https://databricks.com/blog/2018/09/21/how-to-use-mlflow-to-reproduce-results-and-retrain-saved-keras-ml-models.html to see if this guidance helps? This is for retraining model in MLflow. Thanks.

0 Votes 0 ·