Rain Forecasting for Production Capacity Planning at Open Pit Mining

Abstract

Extreme weather events have been one of the biggest challenges in the mining sector. One alternative to mitigate these risks is to improve the prediction of extreme weather events occur. One of the common extreme weather events faced by the mining sector is extreme rainfall. Continuous extreme rainfall can give rise to flooding events, which might disrupt the supply chain and operation of the mining sector. Previously, extreme rainfall prediction is conducted by employing the traditional statistical methods such as linear regression or autoregressive integrated moving average (ARIMA). Those methods result in good accuracy; however, they do not cover some of the assumptions of the data. Along with the development of information technology, advanced method, namely machine learning, is conducted. Thus, this study employed machine learning to predict the rainfall duration in open-pit mining. The predictive models constructed are a feed-forward neural network and an ARIMA model. This study also compared the performance of the neural network model and the ARIMA model by measuring its root mean square (RMSE). Based on the result, the neural network model outperforms the ARIMA model.