Performance Comparison of Text Classification Algorithms and Feature Extraction Techniques on Financial Tweets

This study examines the performance of different machine learning algorithms used in sentiment analysis of financial tweets. The aim is to increase the accuracy of sentiment classification in financial tweets and to determine which feature extraction techniques and models are more effective in this context. In the study, Logistic Regression, Support Vector Machines, Decision Trees and K-Nearest Neighbors algorithms were evaluated using N-grams and TF-IDF feature extraction techniques. Performance evaluation was based on accuracy, precision, sensitivity and f1-score metrics. Simulation results have shown that the TF-IDF feature extraction method excels in all models with the highest accuracy and performance metrics. In particular, Support Vector Classifier and Decision Tree models showed higher performance than other models when TF-IDF was used. Other feature extraction methods (2-gram, 3- gram) generally showed lower performance. This reveals the effectiveness of these techniques and models in sentiment classification of financial tweets. Future studies could extend these findings by using more advanced deep learning models and more diverse datasets. This research makes important contributions to how social media data can be used for financial market analysis and investment decisions.