Automl training process

[maven-ec]

Hey folk,
I'm using automl approach to train some classification text models. I had trained two models with the same data set and got a big difference in the performance metrics between the models. e.g. recall model 1: 0.80, recall model 2: 0.65.
So, I have a question about how the training process works, and if the model depends on a stochastic process (seed).

[anjelab]

For starters, you may read this documentation about precision and recall. Recall tells us, from all the test examples that should have had the label assigned, how many were actually assigned the label. On the other hand, Precision tells us, from all the test examples that were assigned a label, how many actually were supposed to be categorized with that label. Your optimization on either of the two will depend on these use cases. 

If you really are after a comparison of the accuracy of your models, the recommended metric is Average Precision.  It measures how well your model performs across all score thresholds and this is under the precision-recall curve.