NNAA, Sunday Barikui et al, International Journal of Computer Science and Mobile Computing, Vol.11 Issue.5, May- 2022, pg. 92-221 © 2022, IJCSMC All Rights Reserved 92 Available Online at www.ijcsmc.com/journal International Journal of Computer Science and Mobile Computing A Monthly Journal of Computer Science and Information Technology ISSN 2320–088X IMPACT FACTOR: 7.056 IJCSMC, Vol. 11, Issue. 5, May 2022, pg.92 – 221 DEVELOPING AN IMPROVED SENTIMENT ANALYSIS SYSTEMS USING ENSEMBLE OF PREDICTIVE EXPERTS IN FOREIGN EXCHANGE (FOREX) FORECASTING NNAA, Sunday Barikui 1 ; Prof. ASAGBA, Prince O. 2 Department of Computer Science, Ignatius Ajuru University of Education, Port Harcourt, Rivers State, Nigeria 1 Department of Computer Science, University of Port Harcourt, Rivers State, Nigeria 2 1 nnaasunday76@yahoo.com; 2 asagba.prince@uniport.edu.ng DOI: https://doi.org/10.47760/ijcsmc.2022.v11i05.009 Abstract: The study considered the specific risk issues associated with the Foreign Exchange (FOREX) market in Nigeria. Foreign exchange investors are risk takers due to the decisions they often make on their resources for investment. Accurate forecasting of data has often been challenging to investors in the foreign exchange market due to lack of predictive and statistical tools. The mentioned challenge is also as a result of not applying models that accurately predict foreign exchange signals and trends. In this study we developed an Improved Sentiment Analysis Systems using Ensemble of Predictive Experts comprising of neural network, Regression Analysis and Decision Tree was developed. The study adopted the agile methodology due to its ability to iteratively adapt to changing user requirements and the volatile nature of the foreign exchange market. The model was implemented in Python Programming Language and MySQL database as backend. A comparative analysis of the existing and new systems was carried out and the parameters used for the analysis encompassed the methodology adopted, the number of predictions, number of accurate predictions, number of tested records for prediction, and the number of currencies used for the prediction was considered. The results clearly showed that the new system outperformed the existing system with 58 accurate predictions while the former achieved 38 accurate predictions respectively. The new system performed better in term of accuracy on unseen data to reduce generalization error in forecasting of FOREX signal. Keywords: Text Classification, Decision Tree, Ensemble Learning, Regression analysis, neural network