Volume 10 Issue 2 ( June 2023 )

Manufacturing industries have been growing in terms of their market size and their revenue generated in the past decade. With rapidly evolving technologies, accurate forecasting and predictions act as a strong pillar in strategic planning. Predicting the prices of commodities helps in reducing the financial losses incurred while procurement of the inventory. In this paper, we used the deep learning methods to develop some models for predicting the prices. Deep learning is basically a part of the bigger sphere of machine learning and artificial intelligence. It is based on artificial neural networks with representation learning. The models based on deep learning are superior to any machine learning models in predicting the prices, hence our methodology consists of such techniques. This paper is a preliminary attempt to predict the future prices by using hybrid LSTM- Attention-CNN model with the help of the past 10 years data. The proposed model performed significantly better than traditional ARIMA model which has been a benchmark in time-series for a long time.

Keywords: Metal Commodities, Deep learning, CNN, LSTM, Attention, Time series models.