Backtesting Sentiment Signals for Trading: Evaluating the Viability of Alpha Generation from Sentiment Analysis

2025-07-04

AI Summary

This study bridges the gap between sentiment analysis research and quantitative trading practice by empirically evaluating whether news sentiment signals can generate positive alpha in real-world portfolio allocation. Focusing on the Dow Jones Industrial Average (DJIA) 30 constituents, we construct and backtest three sentiment models—two classification-based and one regression-based—where trading decisions are driven directly by continuous sentiment scores, departing from conventional sentence-level classification paradigms. All strategies outperform the buy-and-hold benchmark; the regression-based model achieves the strongest performance, yielding a cumulative return of 50.63% over a 28-month backtest period—significantly surpassing both classification models and the benchmark. To our knowledge, this is the first systematic empirical demonstration of the superiority of regression-style sentiment modeling in investment applications. The findings provide robust evidence and methodological guidance for transitioning sentiment analysis from a natural language processing task to an actionable financial decision-making tool.

Technology Category

• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Machine Learning: Classification and Regression
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Web Mining and Content Analysis: Sentiment analysis and opinion mining
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Sentiment analysis, widely used in product reviews, also impacts financial markets by influencing asset prices through microblogs and news articles. Despite research in sentiment-driven finance, many studies focus on sentence-level classification, overlooking its practical application in trading. This study bridges that gap by evaluating sentiment-based trading strategies for generating positive alpha. We conduct a backtesting analysis using sentiment predictions from three models (two classification and one regression) applied to news articles on Dow Jones 30 stocks, comparing them to the benchmark Buy&Hold strategy. Results show all models produced positive returns, with the regression model achieving the highest return of 50.63% over 28 months, outperforming the benchmark Buy&Hold strategy. This highlights the potential of sentiment in enhancing investment strategies and financial decision-making.

Problem

• Evaluating sentiment-based trading strategies for alpha generation
• Assessing practical application of sentiment analysis in financial markets
• Comparing sentiment models' performance against Buy&Hold benchmark

Innovation

• Backtesting sentiment-based trading strategies
• Comparing three sentiment prediction models
• Regression model achieves highest returns