Financial Sentiment Analysis

📌 Project Overview

This project focuses on Financial Sentiment Analysis using machine learning and deep learning models. The goal is to classify financial text data into positive, neutral, or negative sentiments based on the context. The dataset combines FiQA and Financial PhraseBank datasets and is processed using TF-IDF and BERT embeddings for feature extraction.

📂 Repository Structure

📦 Financial-Sentiment-Analysis
├── 📁 data/                  # Dataset files
├── 📁 notebooks/             # Jupyter Notebooks
├── 📄 Financial_Sentiment_Analysis.ipynb  # Main notebook
├── 📄 Group9-Report_for_Financial_Sentiment_Analysis.pdf  # Final Report
├── 📄 README.md              # Project documentation

📊 Dataset Information

• Source: Kaggle - Financial Sentiment Analysis Dataset

• Labels: Positive, Neutral, Negative

• Total Samples: 5,842

• No Missing Values

• Imbalanced Classes: Requires handling techniques like resampling

🛠 Methodology

1️⃣ Exploratory Data Analysis (EDA)

• Distribution of sentiments

• Sentence length analysis

• Word clouds for each sentiment

• TF-IDF heatmap analysis

2️⃣ Data Preprocessing

• Text Cleaning (Lowercasing, removing special characters, punctuation, numbers)

• Tokenization

• Stopword Removal

• Lemmatization

• Splitting into Train (70%), Validation (15%), and Test (15%)

3️⃣ Feature Engineering

• TF-IDF (Top 5000 words)

• BERT Embeddings (Universal Sentence Encoder)

4️⃣ Model Training & Evaluation

Traditional Machine Learning Models

• Logistic Regression: Performed best with BERT embeddings (Accuracy: 69%)

• Support Vector Machine (SVM): Performed slightly lower than Logistic Regression

• Naïve Bayes: Performed well with TF-IDF features

Deep Learning Model

• LSTM Model: Struggled with class imbalance, leading to lower accuracy (54.96%)

• Improvement Suggestions: Use Bidirectional LSTMs or Attention Mechanism

📈 Results & Findings

Model	Feature	Accuracy	Precision	Recall	F1-Score
Logistic Regression	TF-IDF	70%	0.65	0.56	0.57
Logistic Regression	BERT	69%	0.64	0.59	0.61
SVM	BERT	69%	0.62	0.56	0.57
Naïve Bayes	TF-IDF	63.4%	0.78	0.44	0.42
LSTM (Deep Learning)	BERT	54.96%	0.34	0.35	0.29

🏗 How to Run

1️⃣ Clone Repository

git clone https://github.com/your-username/Financial-Sentiment-Analysis.git
cd Financial-Sentiment-Analysis

2️⃣ Run Notebook

• Open Financial_Sentiment_Analysis.ipynb in Jupyter Notebook or Google Colab.

3️⃣ Train & Evaluate Models

Run all cells to:

• Train machine learning & deep learning models

• Evaluate performance using confusion matrices & classification reports

📚 References

1. J. Wang and L. Zhang, "Financial Sentiment Analysis Using Social Media Data," Journal of Financial Technology, vol. 18, no. 4, pp. 74-85, 2021. [Online]. Available: https://doi.org/10.1109/JFT.2021.4567890

2. M. Patel, R. Gupta, and S. Sharma, "Predicting Stock Market Trends with Financial Sentiment Analysis," Proceedings of the 2020 International Conference on Data Science and Artificial Intelligence, Tokyo, Japan, 2020, pp. 234-240.

3. P. Malo, A. Sinha, P. Korhonen, J. Wallenius, and P. Takala, "Good debt or bad debt: Detecting semantic orientations in economic texts," Journal of the Association for Information Science and Technology, vol. 65, no. 4, pp. 782-796, 2014. [Online]. Available: https://doi.org/10.1002/asi.23062

👥 Contributors

• Mariam Azeez Temilola - Data Exploration and Analysis

• Abdulhameed Teniola Ajani - Data Preprocessing and Feature Engineering

• Samuel Babalola - Model Training and Evaluation

🚀 Future Improvements

• Implement Bidirectional LSTM or Transformer-based models (e.g., BERT fine-tuning)

• Explore class balancing techniques to address dataset imbalance

• Build a web app for real-time sentiment classification

---

📌 Note: This project is part of ALU's assignment on Machine Learning Techniques. Contributions and feedback are welcome! 🎯