Multivariate time series based on deep learning models for mill throughput forecasting using feature selection: an enhanced analysis given grade uncertainty

---

Parlar T., Dimitrakopoulos R.

INTERNATIONAL JOURNAL OF MINING RECLAMATION AND ENVIRONMENT, 2025 (SCI-Expanded, Scopus)

• Yayın Türü: Makale / Tam Makale
• Basım Tarihi: 2025
• Doi Numarası: 10.1080/17480930.2025.2586062
• Dergi Adı: INTERNATIONAL JOURNAL OF MINING RECLAMATION AND ENVIRONMENT
• Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, Environment Index, Geobase, Greenfile, INSPEC
• Hatay Mustafa Kemal Üniversitesi Adresli: Hayır

Özet

This study presents a deep learning framework for mill throughput forecasting in mining operations, combining Gated Recurrent Unit Recurrent Neural Networks (GRU-RNN) with systematic feature selection to enhance accuracy given grade uncertainty. Feature importance is evaluated using permutation importance and SHAP analysis integrated with CatBoost and Gradient Boosting, leveraging multivariate time series data from a gold mining complex. The GRU-RNN model achieved a 15.1% reduction in RMSE and a 4.2% improvement in R2 compared to the baseline. This research introduces mill throughput modelling by establishing a robust framework that provides mining operations with an understanding of the features affecting throughput performance across different model realisations.