AutoML Wizard

The AutoML Wizard at frontend/ml-workbench/src/components/AutoML/ provides a guided 5-step wizard for automated machine learning, from dataset selection through algorithm configuration to training launch.

Components

Wizard Steps

const steps: WizardStep[] = [
  { id: 'dataset', title: 'Select Dataset', icon: TableCellsIcon },
  { id: 'target', title: 'Select Target', icon: SparklesIcon },
  { id: 'algorithm', title: 'Configure', icon: AdjustmentsHorizontalIcon },
  { id: 'training', title: 'Training', icon: CpuChipIcon },
  { id: 'review', title: 'Review', icon: DocumentCheckIcon },
];

Step 1: Dataset Selection

Lists available datasets with metadata:

interface Dataset {
  id: string;
  name: string;
  rows: number;
  columns: number;
  lastUpdated: string;
  size: string;
}

Step 2: Target Selection

Presents columns from the selected dataset with automatic type detection (numeric, categorical, temporal) to suggest appropriate ML task types (regression, classification, time series).

Step 3: Algorithm Configuration

Based on the detected task type, presents relevant algorithms with configurable hyperparameters. Supports "auto" mode where multiple algorithms are compared.

Step 4: Training Parameters

GPU/CPU selection, batch size, epochs, early stopping configuration, and cost estimation.

Step 5: Review and Launch

Summary of all configuration with estimated training time and cost. Launches the training job via the MLApiClient.