Data Reconciliation-Based Hierarchical Fusion of Machine Learning Models DOI Creative Commons
Pál Péter Hanzelik, Alex Kummer, János Abonyi

и другие.

Machine Learning and Knowledge Extraction, Год журнала: 2024, Номер 6(4), С. 2601 - 2617

Опубликована: Ноя. 11, 2024

In the context of hierarchical system modeling, ensuring constraints between different hierarchy levels are met, so, for instance, aggregation satisfied, is essential. However, modelling and forecasting each element independently introduce errors. To mitigate this balance error, it recommended to employ an optimal data reconciliation technique with emphasis on measurement modeling study, three machine learning methods development were investigated. The first method involves no reconciliation, relying solely models built at level. second approach incorporates errors by adjusting measured satisfy constraint, model developed based dataset. third directly fine-tuning predictions prediction model. compared using case studies complexities, namely mineral composition estimation 9 elements, retail sales 14 waste deposition more than 3000 elements. From results conclusion can be drawn that performs best, reliable developed.

Язык: Английский

Data Reconciliation-Based Hierarchical Fusion of Machine Learning Models DOI Creative Commons
Pál Péter Hanzelik, Alex Kummer, János Abonyi

и другие.

Machine Learning and Knowledge Extraction, Год журнала: 2024, Номер 6(4), С. 2601 - 2617

Опубликована: Ноя. 11, 2024

In the context of hierarchical system modeling, ensuring constraints between different hierarchy levels are met, so, for instance, aggregation satisfied, is essential. However, modelling and forecasting each element independently introduce errors. To mitigate this balance error, it recommended to employ an optimal data reconciliation technique with emphasis on measurement modeling study, three machine learning methods development were investigated. The first method involves no reconciliation, relying solely models built at level. second approach incorporates errors by adjusting measured satisfy constraint, model developed based dataset. third directly fine-tuning predictions prediction model. compared using case studies complexities, namely mineral composition estimation 9 elements, retail sales 14 waste deposition more than 3000 elements. From results conclusion can be drawn that performs best, reliable developed.

Язык: Английский

Процитировано

0