TAZI Live for Regression

TAZI Live shows important metrics related to your Business Model in real time. Hence, every chart, table and metric gets refreshed during the run course of your model.

When you choose regression as the problem type, you will see a slightly different TAZI Live page since the metrics related to a regression problem is very different than the classification algorithm.

Counts

You have 3 tabs here. On the counts tab you can see actual and predicted values for each row. On the performance tab you will see a screen similar to below.

Performance

Here you can see instance count and computed metrics such as MSE, MAE, MAPE. These metrics are recalculated as the incoming data is processed since this is how continuous learning works. Also, you can see how a particular metric is computed by hovering your mouse over to its header.

Additionally, if you split your data as a train and test, you will see each result separately in their own tabs.

Model

On the model tab, you can see feature importances and explanations of sample instances. Feature importance is displayed on the left side. The feature explanations are calculated using the Local Interpretable Model-Agnostic Explanations (LIME).

You can filter your data by entering queries into the search field. Also, you can see the actual target and predicted value for each row.

When you click on a particular row, it will expand and you will be able to see its explanations.

For example, in the above screenshot, you can see that MSZoning=0, LotConfig=1, GarageQual=1, CentralAir=1, and LandSlope=0 feature values affected how Tazi makes predictions for this particular row, mostly. For this row, we can observe that the feature value MSZoning=0 affected the house’s price negatively more than LotConfig=1.