Financial Sentiment Analysis Model Using Ebola Optimization Search Algorithm and Bi-Directional Gated Recurrent Unit

In today's fast-paced financial industry, data-driven decision-making and knowledge management are crucial for the growth of Financial Technologies (FinTech) companies. Big Data (BD) performs an essential role in accurate financial analysis by aggregating data from various sources to enhance the efficiency of financial services. The prediction of stock market behavior is a complex task due to the presence of various factors influencing stock prices, such as news and market trends, which affect the accurate Financial Sentiment Analysis (FSA). This research proposes an improved Sigma Point Mutation Strategy-based Ebola Optimization Search Algorithm (SPMS-EOSA) for feature selection. The model is designed to select the most relevant features and enhance the classification accuracy of FSA. The SemEval-2014 dataset is used for model evaluation, and is pre-processed using the tokenization technique, which splits text clusters into smaller, meaningful units called tokens. Feature extraction is performed using the Long-Term Frequency-based Modified Inverse Class Frequency (LTF-MICF) technique, thereby extracting essential features that enhance the model's efficiency. A Bi-directional Gated Recurrent Unit (Bi-GRU) model in FSA captures both contextual and temporal relationships in financial data, helping to identify subtle deviations indicative of potential financial attacks. The proposed method achieves an accuracy of 98.21%, outperforming existing methods such as Convolutional Neural Networks (CNN) combined with Long Short-Term Memory (LSTM).