Abstract: Ore grade estimation is relatively difficult due to the complexity of ore deposit formation process and numerous control factors. Evaluation of ore deposit with low estimation error is crucial in mineral resources development and usage. So far， Kriging， now known as a best estimation method of grade， is based on intrinsic assumption and stationarity about the underlying grade spatial distribution. However， most of ore grade data are spatially sparse， irregularly spaced and have complex distribution， which could result in the Kriging estimation method violating intrinsic assumption and stationarity. This article presented a new method for ore grade estimation based on multi-gene genetic programming and also compared it with ordinary Kriging. The results show that the proposed method makes no assumptions about the spatial distribution of grade data， the condition of implementing ore body grade prediction is simplified， and it can achieve better prediction effect. So， the proposed method can be used to estimate ore grade for complex ore deposit.

Key words: ore grade estimation, multi-gene genetic programming(MGGP), ordinary Kriging, ore deposit prediction, artificial intelligence