Real-Time Forecast of BIST100 Index Under Market Volatility and Uncertainty

Abstract Rapid changes and uncertainties in financial markets necessitate that investors have access to timely and accurate information to make the right decisions. Instant changes and volatility in the markets can negatively affect future forecast success. In this context, this study presents an approach for estimating the real-time values of BIST100, the leading index of Borsa Istanbul. The proposed approach initially predicts the next day's closing price and direction of the BIST100 index, which is a general indicator of the stock market. This is achieved by employing support vector regression, decision trees, random forests, multiple linear regression, polynomial regression, and hybrid models based on ensemble learning techniques. At this stage, five variables of BIST100, 46 technical indicators, and four basic indicators derived from these variables are utilized. To address challenges associated with market volatility and uncertainty, existing algorithms were retrained by incorporating weekly average values alongside daily data. This integration aimed to smooth data variability and improve prediction performance, enabling real-time forecasting using the most effective methods. The findings of the study indicate that the direction and closing values of the BIST100 index can be predicted with improved accuracy when appropriate modeling techniques are employed in conjunction with the optimal parameters and methodologies that are developed to mitigate the impact of instantaneous fluctuations in the data.