Mining capital cost estimation using Support Vector Regression (SVR). (August 2019)

• Record Type: Journal Article
• Authors: Nourali, Hamidreza; Osanloo, Morteza
• Is Part Of: Resources policy. Volume 62(2019)
• Page Start: 527
• Page End: 540
• Publication Date: 2019-08
• Subjects: CAPEX -- Capital cost estimation -- Support Vector Regression (SVR) -- Kernel Ridge Regression (KRR); Mines and mineral resources -- Periodicals; Ressources minérales -- Périodiques; Ressources naturelles -- Gestion -- Périodiques; Environnement -- Politique gouvernementale -- Périodiques; 333.8
• DOI: 10.1016/j.resourpol.2018.10.008
• Languages: English
• ISSNs: 0301-4207

Abstract

Abstract: Determination of Capital Expenditure (CAPEX) is a challenging issue for mine designers. Underestimating the capital cost in mining projects may postpone the construction and accordingly the production phases. In addition, overestimating the capital cost may decrease value of the project. Currently available capital cost estimation models cannot predict mining CAPEX in a reliable range of error. Since, current models are not considering all effective parameters other than capacity, annual ore production and waste stripping, they cannot turn out to a reliable result, although they can be used for a rough estimation of CAPEX. In this paper, to estimate the capital cost of mining projects, a model based on Support Vector Regression (SVR) is developed. To establish this model the technical and economic data of 52 open pit porphyry copper mines were collected. Robust design of this model led to negligible error of estimation of CAPEX anticipation procedure. According to the results, the capability of presented model to estimate the mining CAPEX in a wide range of mining capacity is proved. So, as a whole, with a view of evaluation results, this model can be used as a reliable model for estimating of mining CAPEX. Highlights: In this paper, data of 52 open pit porphyry copper mines have been collected. This study tries to benefit from Support Vector Regression theory to develop a cost estimation model. The research indicates that the performance of the Support Vector Regression (SVR) model is obviously better than Kernel Ridge Regression (KRR) model to predict the mining CAPEX. Four major rules that should be considered in devising a reliable cost model have been presented that these are as follows: 1) Collected data should be related to a specific mineral and specific extraction and processing method, 2) Interval variations of mining capacity and generally scale of mining following a suitable dispersion, 3) Effective factors in determining capital cost should be included in the model, and 4) Due to the complexity of the problem, the appropriate methodology to construct the model is selected.