DOI: 10.11118/ecos2026_038

AI METHODS FOR STOCK-PRICE FORECASTING IN THE INDIAN MARKET: THE EFFECT OF MACROECONOMIC FACTORS

Ankit Tripathi¹, Oldřich Trenz¹

¹ Department of Informatics, Faculty of Business and Economics, Mendel University in Brno, Zemědělská 1, 613 00 Brno, Czech Republic

This article compares four artificial-intelligence forecasting methods (LSTM, SARIMAX, XGBoost and FB-Prophet) on a decade of daily prices of Reliance Industries, one of the largest companies on the Indian stock market. Most published forecasting studies use US, Chinese or European data; the Indian market is less frequently used as a test bed. In the source thesis, a Scopus query for “time series forecasting in India” refined by “financial market” returned 137 documents, while the broader bibliometric dataset contained 2,202 documents. The dataset covers 2 December 2013 to 30 November 2023 (over 2,600 trading days). Each model was first trained on price features alone and then re-evaluated after merging five macroeconomic indicators (gold price, USD/INR, the 10-year government-bond yield, the Wholesale Price Index and the Consumer Price Index). Accuracy was measured by MSE, MAE and the coefficient of determination R². Adding macroeconomic factors did not help every model. XGBoost improved across all three metrics (MSE 39.35 to 24.58) and was the only method that gained without producing implausibly perfect numbers; LSTM improved only with the larger 90% training split, while FB-Prophet overfitted badly (MSE 11.17 to 1,912.35). The results are specific to one company and a retrospective forecasting experiment on historical data; they do not represent live trading performance.

Keywords: time series forecasting, LSTM, XGBoost, SARIMAX, Indian stock market, macroeconomic factors

pages: 348-355, online: 2026