IL3.3A - SPPF High-Accuracy Spare-Parts Forecasting Platform
A Docker microservices platform for forecasting spare-parts consumption. It includes a forecast database, REST API, periodic hyperparameter retraining and theoretical components validated in international scientific publications, and is fed by SAP/Bosch APIs.
-
Main Gain(s): Spare-parts demand forecasting, Periodic model retraining, REST API integration, Support for long-term performance monitoring
-
Core Contributions: (concept) E. Rocha; (implementation) J. Sousa, F. Vieira (testing) J. Sousa (server integration) J. Sousa
Use Cases:
-
SAP/Bosch spare-parts consumption time series and real-time data feeds via REST API
-
Start TRL: 5 - …
-
Final TRL: 7 - …
-
Main Features
-
Sparse time-series forecasting: Predicts spare parts demand even when historical data is limited, irregular, or highly sparse (common for sales of spare parts).
-
Zero-inflated demand modeling: Handles logistics datasets with many zero-demand or zero-capacity observations, where conventional forecasting models often underperform.
-
AI-based model selection: Uses proprietary meta-ranking algorithms to identify which forecasting models are most suitable for each demand pattern.
-
AI Traceability: All models and trains are registered for future audit.
Images and Videos
| Product Demonstration |