Nickel Price Prediction Using Bi-Directional LSTM (Bi-LSTM) and Attention Bi-Directional LSTM Network (At-Bi-LSTM)

The fluctuatives trend of nickel price possess significant challenges to nickel mining and smelting company, that possess high financial risk due to its capital intensiveness trait. Providing decision maker with robust price prediction model, can help the business leader to create more relevant and reasonable business decision, either in daily operations or long term investment. Such method to aid price prediction has been employed, one of which is using deep learning method such as Long Short Term Memory (LSTM). This model, promise better predictive capability compared to more traditional model, due to its capability in capturing more complex pattern in the data. However, in more chaotic time - series trend, this model also prone to inaccuracy, due to difficulties to capture more random fluctuation. This research aims to investigate price prediction model using bi-directional LSTM, further enhanced with attention mechanism, which recently gained more attention per se because it allows the model to focus on specific trends in the data, that helps explained short term and long-term fluctuation of the model. The results of this research, show that Bi-LSTM model can outperform LSTM model obtained from other research, while attention Bi-LSTM still outperformed by both Bi-LSTM and the previous model. Further research is needed to confirm the methods capability in dealing with more complex and fluctuative time series trend.