Skip to main content

EvoML Core Features

EvoML provides a comprehensive suite of tools for end-to-end machine learning workflows. Below is a breakdown of its core capabilities:

1. Data Upload and Processing​

Data sources:​

  • Supports structured formats (CSV, Excel, JSON).
  • Handles unstructured data formats (text, images).
  • Connects to databases (SQL, NoSQL).
  • Integrates with cloud storage (S3, Azure Blob).
  • Supports FTP connections.

Automated data preparation:​

  • Automated data cleaning and preprocessing.
  • Missing value detection and handling.
  • Outlier detection.
  • Data type inference and conversion.

Feature analysis:​

  • Statistical insights (mean, median, variance).
  • Pattern detection and analysis.
  • Visualization tools:
    • Histograms
    • Density plots
    • Box plots
    • Correlation matrices

2. Model Development​

Hyperparameter Tuning:​
  • Tree-Structured Parzen Estimator.
  • Non-Dominated Sorting Genetic Algorithm II.
  • Random.
Model generation:​
  • Enable manual setting of model parameters.
  • Automated/manual model selection.
  • Cross-validation strategies.
  • Automated reporting.

3. Feature Engineering​

Impute, Encode and Scale:​

  • Multiple imputation strategies.
  • Categorical encoding methods.
  • Numeric scaling techniques.
  • Standardization and normalization.
  • Generate embeddings from text.

Feature Generation:​

  • Automated feature creation.
  • Interaction term generation.
  • Polynomial feature expansion.

Feature Selection:​

  • Minimum redundancy maximum relevance.
  • Feature importance ranking.
  • Correlation-based selection.
  • Dimensionality reduction techniques.

4. Timeseries Handling​

Model Validation:​

  • Time-based cross-validation.
  • Sliding Window.
  • Expanding Window.

Feature Engineering:​

  • Time-based feature generation.
  • Lag generation.
  • Rolling windows.

5. Multi-objective Optimization​

  • User-defined objective function.
  • Up to three simultaneous optimization criteria.
  • Pareto optimization.

6. Evaluation​

Model Evaluation:​
  • Classification/regression metrics.
  • Custom evaluation criteria.
  • Visualisations.
Explainability:​
  • Feature importance analysis.
  • SHAP values.
  • Partial dependence plots.
  • Model-agnostic explanations.

7. EvoML Client​

Code Interface:​
  • Python client library.
  • RESTful API integration.
  • Batch processing support.
  • Real-time prediction capabilities.
  • Model retraining.
  • Extendable code interface.

8. Deployment​

Pipeline Generation:​
  • Production-ready code generation.
  • Environment management.
  • Dependency handling.

Model deployment:​

  • Automated API generation.
  • Docker containerization.
  • Model monitoring capabilities.
  • Integration with MLOps tools.