Baselines
Baseline Evaluation Metrics
On your predictive query's evaluation page, Kumo provides comparisons of your model versus some heuristic baselines for temporal binary classification, link prediction, and regression tasks. This allows you to compare the results of your Kumo model to results from very simple approaches.
Baselines are computed on a per-entity level (i.e., by using the past label as a forecast for regression, or by recommending past user behavior for link prediction).
Note: Baselines are not supported when
timeframe_stepisNone.
Baselines Per Task Type
The following are the current baselines supported by Kumo, organized prediction task type:
| Task Type | Temporal Problem |
|---|---|
| Link Prediction | Per Entity Frequency: Predicting each entities with the most popular ranked list of target values for this entity in all previous splits |
| Regression/Logistic Regression | Per Entity Mean: (personalized average) Predicting each entity with the average of all target values for this entity in all previous splits Per Entity last 1 label: Predicting each entity with the target value in the previous time frames |
| Binary Classification | Per Entity Mean: Predicting each entities with the average of all target values for this entity in all previous splits Per Entity last 1 label: Predicting each entity with the target value in the previous time frame |
Updated 8 months ago