Sentiment-Weighted Fusion of Financial News and Technical Indicators for Short-Horizon Equity Trend Classification: An Evidence-Informed Research Framework

Abstract

Short-horizon equity trend prediction remains difficult because price formation reflects both historical market behaviour and rapidly changing information flows. Technical indicators condense past open-high-low-close-volume data into interpretable measures of trend, momentum, volatility and participation. At the same time, financial news sentiment captures the market relevance of textual disclosures, macroeconomic updates and company-specific events. This paper develops a sentiment-weighted fusion framework that combines transformer-derived financial news sentiment with selected technical indicators for next-day equity trend classification. The proposed design uses a finance-domain transformer, such as FinBERT, to transform news headlines or short articles into positive, negative and neutral sentiment probabilities; these outputs are then aggregated by trading date and merged with engineered indicators including moving averages, relative strength index, moving average convergence divergence, Bollinger Bands, daily return, volume change and historical volatility. Instead of presenting unverified or simulated accuracy values, the manuscript positions the contribution as an evidence-informed methodological framework supported by established studies on public mood, LSTM-based market forecasting, transformer-based sentiment classification and hybrid sentiment-technical prediction. The paper specifies the research gap, hypotheses, input variables, feature construction, validation design, evaluation metrics and reproducibility requirements for future empirical implementation. The framework is intended to support a narrower publishable subset of a broader multimodal financial market prediction research agenda while maintaining transparency about the need for original dataset validation before final performance claims are made.