Sentiment_Analysis_Stock_Tweets_FineTuned

Model Overview

Model Name: vidhivaish03/Sentiment_Analysis_Stock_Tweets_FineTuned
Model Type: Transformer-based, fine-tuned
Base Model: ahmedrachid/FinancialBERT-Sentiment-Analysis
Language: English
Task: Sentiment Analysis (Bearish, Bullish, Neutral)

Dataset

Dataset Name: zeroshot/twitter-financial-news-sentiment
Description: The dataset consists of financial news tweets annotated with three sentiment labels: Bearish, Bullish, and Neutral. It includes 9,543 training samples and 2,388 validation samples.
Data Structure:

• Training Set: 9,543 samples
• Validation Set: 2,388 samples

Training Details

Preprocessing: Text cleaning (removal of URLs, special characters), tokenization, stop word removal, lemmatization, and vectorization using Word2Vec embeddings.
Training Framework: Hugging Face's Transformers library
Fine-Tuning Epochs: 5
Optimizer: AdamW
Learning Rate: 2e-5

Model Performance

Classification Report

The classification report provides a detailed performance overview of the fine-tuned FinancialBERT model on the validation set, evaluating its ability to classify stock-related tweets into Bearish, Bullish, and Neutral sentiments.

               Precision Recall    F1-Score  Support
 Bearish       0.83      0.69      0.76       285
 Bullish       0.75      0.83      0.79       391
 Neutral       0.91      0.91      0.91      1233

accuracy                           0.86      1909
macro avg      0.83      0.81      0.82      1909
weighted avg   0.86      0.86      0.86      1909

Use Cases

• Investor Sentiment Analysis: To gauge market sentiment from social media posts and financial news, aiding investment decisions.
• Market Trend Prediction: To anticipate stock movements based on aggregate sentiment analysis from a variety of sources.
• Financial News Monitoring: To automatically categorize financial news articles and tweets into sentiment categories, streamlining information gathering for analysts.

Limitations

• Domain Specificity: The model is fine-tuned specifically for financial news tweets, and may not generalize well to other domains.
• Context Understanding: While FinancialBERT captures financial context well, it might not interpret nuanced sentiments or sarcasm effectively.
• Data Bias: The model's performance is contingent on the quality and representativeness of the training dataset, which might include biases inherent in social media discussions.

Future Work

• Domain Expansion: Fine-tune additional models for other domains such as general news or consumer reviews.
• Multi-Modal Analysis: Integrate sentiment analysis with other data types like images or financial indicators for more comprehensive insights.
• Enhanced Interpretability: Develop methods to better interpret and explain model predictions to non-technical stakeholders.

Deployment

Hugging Face Hub: The model is available on Hugging Face Hub for easy integration and deployment.
Inference: The model can be accessed via RESTful APIs or integrated into financial sentiment analysis pipelines for real-time predictions.

Repository Link: vidhivaish03/Sentiment_Analysis_Stock_Tweets_FineTuned

---

This model card summarizes the key aspects of vidhivaish03/Sentiment_Analysis_Stock_Tweets_FineTuned, providing insights into its training, performance, and applications.