Geopolitical Sentiment as a Leading Indicator: A Hybrid Analytics Approach to Forecasting Oil Volatility and Emerging Market Vulnerability (2015–2025)

Abstract

The study logically explores the fluid nature of geopolitical stress on international oil prices, defense performance markets, and emerging market currencies with a combination anal analytical pipeline. The paper considers GARCH modeling, event study method, sentiment analysis, and machine learning (XGBoost) and applies it to the data that runs between 2015 and 2025 taking into consideration such key flashpoints as the Russia, Ukraine crisis, the Iran, and Israel conflict, or the repeat confrontation in the Red Sea. A novel set of 30+ conflict events manually mastered was supplemented with the sentiment values computed with the TextBlob and indicated in domains. The results have shown a statistically significant evidence of volatility clustering around conflict days in the oil prices, a post-event average standardized increase of 6.8 percentage points in oil prices and directional consistency in the exchange rate of USD/PKR. The model, which predicted the accuracy rate of the model to be the highest was XGBoost model, which was built based on sentiment data as well as the oil return data in order to predict the accuracy rate of the movements in PKR at 59% degree of accuracy. There is good news despite this modest degree of accuracy: this is already not bad in financial prediction other than in the prediction of small gains that are operationally equivalent to greater chance (50%). Interactive dashboards established in Power BI and Dash are used to support the real-time situation analysis and policy modeling. The result contributes to the current body of literature of the geopolitical risk priced as well as demonstrating the need of considering econometric models and the integration of AI-facilitated structures to support financial risk forecasting.