NLP Stock Sentiment Analysis — AI for Market Signals

Description

NLP Stock Sentiment Analysis — AI for Market Signals is a transformer-based machine learning model that applies Natural Language Processing (NLP) to financial news to automatically classify sentiment and generate weekly market insights.

Problem Context:
Financial analysts face challenges tracking investor sentiment across the growing volume of daily market news. Manual analysis is time-consuming and often biased, making it difficult to detect early sentiment shifts that influence stock prices and trading strategies.

Solution Overview:
This AI-driven system processes unstructured news text, predicts sentiment polarity (positive, neutral, or negative), and aggregates sentiment trends weekly. The results provide actionable market indicators that enable investment teams to anticipate movements, refine forecasts, and support data-driven decision-making.

Model Details:

• Framework: PyTorch / Hugging Face Transformers

• Base Model: ProsusAI/FinBERT (finance-tuned BERT)

• Pipeline Tag: text-classification

• Dataset: Historical daily news paired with OHLCV stock data

• Input: News text, Date, and OHLCV columns

• Output: Sentiment label with confidence and weekly rollups

• Evaluation Metrics: Accuracy, weighted Precision, Recall, F1-score, and Confusion Matrix

Why It Matters:

• Converts unstructured news into measurable trading signals

• Enables early detection of market sentiment shifts

• Scales efficiently across large text volumes

• Supports predictive modeling and investment risk analysis

📘 Model Page: Hugging Face – Stock Market News Sentiment Analysis
📄 Research Documentation: DOI: 10.13140/RG.2.2.15049.20320

This preprint is published for open research and reproducibility under the Creative Commons Attribution 4.0 (CC BY 4.0) License.