A web-based application that combines machine learning and explainable AI to predict, explain, and improve operational performance. Users can upload datasets, train predictive models, perform root cause analysis, and obtain actionable recommendations through interactive SHAP, PDP, and ICE visualizations. The platform supports both regression and classification problems, integrates real-time weather data, and provides prescriptive guidance for logistics and port operations.
-
Main Gain(s): Faster identification of the root causes behind operational inefficiencies, delays, and performance issues; Improved decision-making through transparent and trustworthy recommendations; Automated generation of actionable prescriptions
-
Use Case(s): Something ….
-
Start TRL: 2 - (evidences)
-
Final TRL: 6 - (evidences)
-
Main Contributions: (theoretics) Mariano, E. Rocha; (implementation) Mariano
Main Features
- Upload and analyze operational datasets in CSV, Excel, JSON, and Parquet formats.
- Interactive data exploration with descriptive statistics and variable distribution analysis.
- Data preprocessing tools, including feature selection, categorical encoding, and creation of derived variables through arithmetic operations.
- Support for both regression and classification machine learning problems.
- Training, evaluation, and comparison of multiple machine learning models, including XGBoost, Random Forest, SVR, KNN, Ridge, Lasso, and HistGradientBoosting.
- Automated hyperparameter tuning for improved predictive performance.
- Explainable AI (XAI) capabilities using SHAP for feature importance analysis, force plots, waterfall plots, and model interpretation.
- Interactive Individual Conditional Expectation (ICE) analysis to understand variable effects on individual predictions.
- One-dimensional, two-dimensional, and three-dimensional Partial Dependence Plot (PDP) visualizations.
- Automated root cause analysis to identify the key factors driving operational outcomes.
- Prescriptive analytics that generate actionable recommendations for improving performance and reducing delays.
- Counterfactual analysis to determine how specific variables should be adjusted to achieve target outcomes.
- Integration of real-time weather information through external APIs to contextualize operational conditions.
- Secure multi-user web platform with authentication, encrypted data storage, and REST API access.
- Containerized deployment using Docker and Docker Compose for reproducibility and scalability.
- Extensible architecture applicable beyond maritime logistics to other industrial and operational domains.
Videos
| Tool UI Demonstration |