Prediction of Water Quality Index using Deep Learning in Mining Company

The quality of mine water at the mining company XYZ already has an Internet-of-things-based measuring tool. The quality of mine water is a standard that must be complied by the company to the regulator. The impact of passing the quality standard is the temporary suspension of mining activity. An effective process for implementing mine water quality is by predicting mine water quality based on historical data. The data is obtained every two minutes in areas in the company. Forecasting using ARIMA and Support Vector Regression (SVR) have used for years. In recent years, Recurrent Neural Networks (RNN) have shown more correct prediction results than ARIMA. Long Short-Term Memory (LSTM) is a RNN model that uses past data (long term) to predict current data (short term). The results of this research are comparison model of water quality using ARIMA, SVR, and LSTM. It shown that LSTM algorithm gave the best result with lower error. The model from the LSTM method can be used to make predictions such seven days prediction of the pH value in next day whether it is following the rules or needs to be controlled. Because of this preventive maintenance, the company will not be penalized.