%0 Journal Article %T Comparison between the performance of four metaheuristic algorithms in training a multilayer perceptron machine for gold grade estimation %J International Journal of Mining and Geo-Engineering %I University of Tehran %Z 2345-6930 %A Alimoradi, Andisheh %A Hajkarimian, Hossein %A Hemati Ahooi, Hamidreza %A Salsabili, Mohammad %D 2022 %\ 06/01/2022 %V 56 %N 2 %P 97-105 %! Comparison between the performance of four metaheuristic algorithms in training a multilayer perceptron machine for gold grade estimation %K Multilayer Perceptron %K metaheuristic machine learning %K grade estimation %K inverse modeling %K Optimization %R 10.22059/ijmge.2021.314154.594880 %X Reserve evaluation is a very difficult and complex process. The most important and yet most challenging part of this process is grade estimation. Its difficulty derived from challenges in obtaining required data from the deposit by drilling boreholes, which is a very time consuming and costly act itself. Classic methods which are used to model the deposit are based on some preliminary assumptions about reserve continuity and grade spatial distribution which are not true about all kind of reserves. In this paper, a multilayer perceptron (MLP) artificial neural network (ANN) is applied to solve the problem of ore grade estimation of highly sparse data from zarshouran gold deposit in Iran. The network is trained using four metaheuristic algorithms in separate stages for each algorithm. These algorithms are artificial bee colony (ABC), genetic algorithm (GA), imperialist competitive algorithm (ICA) and particle swarm optimization (PSO). The accuracy of predictions obtained from each algorithm in each stage of experiments were compared with real gold grade values. We used unskillful value to check the accuracy and stability of each network. Results showed that the network trained with ABC algorithm outperforms other networks that trained with other algorithms in all stages having least unskillful value of 13.91 for validation data. Therefore, it can be more suitable for solving the problem of predicting ore grade values using highly sparse data. %U https://ijmge.ut.ac.ir/article_86273_4001a6aca449ef3ab7efa45d6adcfd46.pdf