<![CDATA[Stock price forecasting is one of the analytical approaches used by capital market participants to identify future price movement patterns. This study evaluates the performance of the Support Vector Regression (SVR) model in predicting the stock price of United Tractors (UNTR) by optimizing the model’s parameters using two metaheuristic algorithms. The selection of SVR is based on its ability to handle nonlinear regression problems through the use of the Radial Basis Function (RBF) kernel. The parameter optimization of SVR is carried out using the Fruit Fly Optimization Algorithm (FOA), an algorithm inspired by the olfactory and visual system of fruit flies in locating food sources. The advantage of FOA lies in its computational simplicity and fast convergence speed. This study also implements Particle Swarm Optimization (PSO) for comparison purposes. This algorithm adopts a collaborative mechanism among particles in the search space, inspired by the flocking behavior of birds. The stock price data used in this study, covering the period from January 2020 to December 2023, was obtained from Yahoo Finance (https://finance.yahoo.com). The results show that SVR-FOA yields a parameter combination of C = 1000, gamma = 0.9182, and epsilon = 0.9997, while SVR-PSO produces a different configuration, namely C = 1000, gamma = 0.0001, and epsilon = 1. Accuracy evaluation using Mean Absolute Percentage Error (MAPE) indicates that the SVR-PSO model achieves a MAPE of 2.3164%, suggesting a relatively low prediction error. SVR-FOA yields a MAPE of 5.8727%, which is still within the acceptable tolerance range for financial data. While this study focuses on a single stock and uses only historical closing prices, its results provide a strong baseline for applying SVR with metaheuristic optimization in financial forecasting. This research contributes by presenting a direct comparative analysis of FOA and PSO for SVR parameter tuning in an emerging market context, offering practical insights for investors and researchers seeking robust forecasting models.]]>
