Sentiment-Driven Trading Strategy

Project Overview

The project focused on integrating Natural Language Processing (NLP), machine learning, and quantitative trading concepts into a sentiment-driven trading strategy capable of analysing financial news headlines and converting market sentiment into actionable trading signals.

The system processed financial news data, extracted entities using NLP techniques, classified sentiment, generated trading signals, and evaluated portfolio performance using historical market data and backtesting frameworks.

Project Objective

Traditional quantitative strategies often rely only on historical price movements and technical indicators. This project aimed to incorporate market sentiment and financial news analysis into trading decision-making frameworks to improve signal generation and market interpretation.

The primary objective was to evaluate whether sentiment extracted from financial news headlines could improve directional market predictions and enhance trading strategy performance.

System Architecture

• Automated financial news headline collection and preprocessing
• Named Entity Recognition (NER) for company/entity extraction
• Sentiment classification using machine learning models
• Event classification for contextual news understanding
• Trading signal generation based on sentiment output
• Historical market-data integration using yFinance API
• Backtesting engine for strategy simulation
• Interactive Streamlit dashboard for portfolio visualization

NLP & Machine Learning Framework

The NLP pipeline utilised Named Entity Recognition (NER) to identify companies and organisations mentioned within financial headlines. Sentiment classification models categorised headlines into Positive, Negative, and Neutral classes using TF-IDF vectorization and machine learning classification techniques.

Event classification further categorised news into earnings announcements, product launches, regulatory developments, and market updates to improve contextual interpretation of market-moving information.

Trading Strategy Logic

• Positive sentiment → BUY signal
• Negative sentiment → SELL signal
• Neutral sentiment → HOLD signal

Generated signals were tested against historical stock price data through a simulated trading engine to evaluate profitability, portfolio growth, and signal effectiveness.

Performance & Dashboard Analytics

An interactive dashboard was developed using Streamlit to visualise strategy performance, sentiment distributions, signal frequencies, portfolio value growth, and detailed trade-level data.

Backtesting analysis generated approximately 5.18% portfolio return during testing with dynamic portfolio tracking, sentiment monitoring, and signal-performance evaluation.

Risk-adjusted evaluation included metrics such as Sharpe Ratio and Maximum Drawdown to assess portfolio stability and trading consistency.

Key Findings & Insights

• Financial news sentiment can significantly influence short-term trading behaviour and market reactions.
• NLP techniques improve contextual understanding of financial events and company-specific developments.
• Combining sentiment analysis with quantitative frameworks creates more adaptive trading strategies compared to purely technical systems.
• Interactive dashboards improve interpretability of strategy performance and signal tracking.
• Market sentiment analytics can enhance decision-making frameworks for algorithmic trading systems.

Role & Responsibilities

• Built NLP-driven sentiment analysis framework
• Developed trading signal generation logic
• Implemented machine learning sentiment classification models
• Performed financial news preprocessing and entity extraction
• Created historical backtesting simulation engine
• Designed interactive Streamlit dashboard
• Analysed trading strategy performance metrics
• Integrated yFinance API market-data workflows

Skills & Tools Used

Key Learnings

Developed practical understanding of how quantitative finance, NLP, and machine learning techniques can be integrated into market intelligence and trading systems.

The project strengthened capabilities in data-driven trading analysis, financial technology workflows, Python-based analytics, sentiment modelling, and quantitative strategy evaluation.