Sentiment Analysis For Forecast The Gold Price Using Hybrid CNN-LSTM Model

Abdulaziz, Khaled Gamal Thabit (2024) Sentiment Analysis For Forecast The Gold Price Using Hybrid CNN-LSTM Model. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003222020-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (3MB) | Request a copy

Abstract

Gold holds a unique place in human history perhaps the importance lies in its culture significant across the civilizations, However, gold value often stays constant or occasionally rises when other assets increasing or decreasing. Therefore, many researchers and analytics has been conducted many studies to forecast the gold price using a different forecasting models based on macroeconomic indicators and other studies used sentiment analysis approach to forecast the gold price while few studies investigate combining both structure and unstructured data to forecast gold price. Consequently, this study aimed to improve the gold price forecast in Saudi Arabia by combining historical market data such as Petrochemical price(X1), Oil price(X2), Exchange rate(X3) and Inflation(X4) with sentiment analysis of Arabic tweets(X5) as unstructured data. However, for sentiment analysis to classify the tweets as positive or negative Soft Voting Classifier has been used and evaluated by F-1score matric. To forecast the gold price based on structure and unstructured data the proposed hybrid CNN-LSTM model used and evaluated using two matric MAPE and RMSE. The main finding in this research are the Soft Voting Classifier showed a good performance in classifying the tweets as positive and negative with F1 score 0.92. Similarly, incorporating tweet scores into the CNN-LSTM model for forecasting gold prices significantly improves its performance. The model with tweet scores demonstrates a lower (MAPE) of 1.38% compared to 2.65% without tweet scores, Additionally, the (RMSE) is reduced from 1.28 to 0.81, showing closer alignment with actual values. The results highlight the positive impact of integrating sentiment analysis from Twitter on forecasting accuracy. However, this research provided an idea that sentiment analysis has to be considered in forecasting a gold prices which will enhance the investors and analytics decision.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	historical data, sentiment analysis, forecasting gold price, CNN-LSTM
Subjects:	Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.9 Computer algorithms. Virtual Reality. Computer simulation. Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	khaled gamal thabit abdulaziz
Date Deposited:	05 Nov 2024 05:38
Last Modified:	05 Nov 2024 05:38
URI:	http://repository.its.ac.id/id/eprint/115782

Actions (login required)

View Item