Trial

Trial Workflow

The platform guides users through the ML process using a structured workflow:

1. ML Task: The platform automatically determines the ML task based on the data, with options for user adjustments.
2. Objective function: enables the optimisation of models based on various criteria.
3. Data Splitting: Configure how the dataset is split into training and testing sets.
4. Model Validation: Supports validation strategies like k-fold cross-validation for classical models and sliding window validation for time series models to assess generalisation and prevent overfitting.
5. Data Imbalance (classification only): Detect and handle class imbalance using techniques like oversampling, undersampling, or synthetic data generation (e.g., SMOTE).
6. Timeseries: Enables time series analysis, allowing users to select the index column, horizon, and context window.
7. Model Hub (NLP only): Provides access to pre-trained natural language processing (NLP) models for sequence classification and regression tasks.
8. Feature Engineering: Configures encoders, scalers and impute strategies, as well as feature selection, feature generation and dimensionality reduction techniques.
9. Model Selection: Includes a range of models, hyperparameter optimisation, and budget and tuning options.
10. Results Analysis: Provides performance evaluation through various metrics and visualisations.
11. Deployment: The platform generates optimised, production-ready code for the selected model.

Run Trial

1. Create New Trial.
2. Select the dataset, specify the target column, and choose the ML task.
3. Set the objective function for model performance and select the optimiser for hyperparameter optimisation.
4. Configure Splitting, Model Validation and Data Imbalance (for classification).
5. Set Feature Engineering options: choose Feature Selection, Feature Generation and Dimensionality Reduction.
6. Choose Models and tune parameters (if necessary).
7. Click Start on top right corner to start the trial. To cancel the trial, click the Cancel button on top right.
8. Evaluate model performance.
9. Download model and deploy the pipeline in your workflow.
10. Generate report.