UiPath Pipeline Full Pipeline run

Help Automation Ops
1

While running the “Full Pipeline” what will happen on the below scenarios ?

  1. The Evaluate and Train data are same
  2. The Evaluate and Train data are different
2

Hi @Ritaman_Baral,

In scenario 1, the model validation cannot be tursted as the data used to validate is part of the training data. In short model cannot generalise and becomes a rote learning system. “I know what I studied nothing else”

In scenario 2, the model validation can be trusted ( might not be upto the standards you set) but this is the approach which is best practice in any ML pipeline. “I can generalise my learning to novel scenarios”

Training validation and test sets are usually used to model, validate and verify any machine learning model.

3

It will give higher accuracy as the same data model had used for training as well. Confidence score cannot be reliable.

This will give you realistic Confidence score as you separated training & evaluation data.

Thanks,
Ashok :slight_smile:

4

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.