Recent Advances in Deformable Graph Convolutional Networks with NLP Based Social Sentimental Data for Enhanced Stock Price Predictions: A Systematic Review

Abstract

Financial markets are complex, dynamic systems characterized by nonlinear interactions, high-dimensional dependencies, and strong influence from external information such as news and social media. Traditional stock prediction models, including statistical and shallow machine learning approaches, often struggle to capture these complexities, especially when integrating unstructured textual data. This systematic review explores the integration of Deformable Graph Convolutional Networks (DGCNs) with Natural Language Processing (NLP)-based sentiment analysis for improved stock price prediction. DGCNs enhance traditional graph neural networks by introducing adaptive receptive fields and flexible spatial sampling, enabling better modeling of dynamic financial relationships among stocks, sectors, and macroeconomic variables. Simultaneously, transformer-based NLP models such as BERT and FinBERT extract contextual sentiment from diverse textual sources, including financial news, social media, and earnings reports. These sentiment features are incorporated into graph structures as node and edge attributes, enabling a multimodal framework that combines structured financial data with unstructured information. Empirical findings across global stock markets and cryptocurrency datasets indicate that such hybrid approaches outperform conventional and single-modality models. However, challenges persist in data quality, sentiment noise, and temporal alignment. This review highlights key advancements, identifies research gaps, and outlines future directions for developing robust, interpretable, and intelligent financial forecasting systems.