Beyond Sequential Prediction: Learning Financial Market Dynamics in Volatile and Non-Stationary Environments through Sentiment-Conditioned Generative Modelling

TL;DR

This paper introduces a hybrid generative model combining GANs and NLP sentiment analysis to improve time-series forecasting in volatile, non-stationary financial environments by leveraging both numerical and textual data.

Contribution

It presents a novel sentiment-conditioned generative modeling approach that jointly captures market dynamics and exogenous textual information for more robust predictions.

Findings

Demonstrates improved forecasting accuracy in volatile markets.

Shows effectiveness of combining GANs with sentiment analysis.

Highlights potential for hybrid models in non-stationary environments.

Abstract

The problem of time-series forecasting in non-stationary and complex environments is a challenging task in machine learning, especially with heterogeneous numerical and textual data present. Traditional statistical models like AutoRegressive Integrated Moving Average (ARIMA) are based on the assumptions of linearity and stationarity, whereas recurrent neural networks like Long Short-Term Memory (LSTM) models do not necessarily represent distributional properties in highly volatile settings. This paper proposes a hybrid model that combines Generative Adversarial Networks (GANs) with Natural Language Processing (NLP)-based sentiment analysis to enable sentiment-conditioned time-series prediction. The model integrates adversarial learning on numerical sequences with contextual sentiment representations derived from unstructured text, enabling them to be jointly modelled to capture temporal dynamics and exogenous information. These results demonstrate the promise of hybrid generative and language-aware methods to enhance prediction robustness in non-stationary environments.