Deep Learning and Natural Language Processing Integrated Gold Price Forecasting

This paper aims to developing a model based on Deep Learning (DL) and Natural Language Processing (NLP) to improve gold price prediction. The analysis and the forecasting are done using different methods such as; Recurrent Neural Network (RNN), Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional Long Short Term Memory (Bi-LSTM). Such models use the past price of gold and sentiment analysis of news articles as the basis for deriving the results.. Raw data compiled for the analysis covers day-wise gold prices for the period 2002-2024 and headlines, with an emphasis on gold. The results make it evident that the Bi-LSTM model demonstrates the lowest Mean Absolute Error (MAE) of 1. It achieves a total accuracy of 97% with a Mean Squared Error (MSE) of 2. This performance is much better than that shown by RNN and CNN traditional models, which presented high errors and low accuracy measures. Combining sentiment analysis with Named Entity Recognition (NER) also enhanced the prediction effectiveness, improving the sources’ evaluation and identifying more factors affecting the gold price that led to more accurate predictions.