Create a Tabular Experiment
Create a table-data-based training experiment. Creation follows a 4-step wizard — Select domain → Upload data → Column setup → Model Setup — moving with the Next button at the bottom of each step (enabled once the step's requirements are met).
Step 1. Select domain (Domain & Use case)
Click Create experiment, then:
- Select an industry domain card (Finance, Marketing, E-commerce, Insurance, Telecom, Manufacturing, Security, Healthcare, Real Estate, General).
- In the Select a use case section that appears, choose a use case card. Next becomes active.
Step 2. Upload data (Dataset selection)
Select or upload a dataset for training.
| Item | Description |
|---|---|
| Recommended datasets | Datasets recommended for the domain/use case |
| My datasets | User datasets uploaded to the project |
| Search | Find datasets by name, etc. |
Step 3. Column setup
| Item | Description |
|---|---|
| Target variable | Column the model predicts (required) |
| Exclude Columns | Columns to exclude from training (optional) |
You must select a Target variable to enable Next.
Step 4. Model Setup (Validation strategy & Experiment config)
Validation strategy
| Item | Description |
|---|---|
| Split Method | Holdout · Cross Validation · Stratified K-Fold · Time Series |
| Train Data Ratio | Train / Validation / Test ratio (slider) |
| Evaluation Metric | AUC · F1 · ACCURACY · LOGLOSS · MCC |
Experiment config
| Item | Required | Description |
|---|---|---|
| Experiment Name | ✓ | Name identifying the experiment |
| Local queue selection | ✓ | Local queue to run training |
| Optimization Metric | ✓ | Metric to optimize during model search |
Training models can be Auto Select or chosen among LightGBM · XGBoost · CatBoost · Logistic Regression · Random Forest. Review your settings and click Create to create and start the experiment.