ANALYSIS OF THE EFFECT OF SVR KERNEL PARAMETERS VARIATION & DATASET RATIO IN OPTIMIZING BANK STOCK PRICE PREDICTION

• Main Author: Lokheswara Renanda, Enggar
• Format: Final Project
• Language: Indonesia
• Subjects: Fisika
• Online Access: https://digilib.itb.ac.id/gdl/view/46023
• Institution: Institut Teknologi Bandung

Physics which aims to explain a phenomenon by modeling and theories, in its development seeks to model complex systems. The field of physics in complex systems that is rapidly developing is econophysics that tries to model economic systems, especially capital markets. In the capital market, stock prices that change every time in an uncertain manner cause a risk when investing. This risk can be anticipated by doing technical analysis, which is an analysis that involves search of patterns from historical data. One method that can be used to model and predict stock prices is the SVR method. This paper aims to analyze the effect of parameter variations and dataset ratios on the prediction results. The experiment was carried out by modeling the stock price with parameter values and dataset ratios that varied to be analyzed. Modeling with variations is done using the Kernel function and data frequency which provides the best modeling results on BCA and BRI shares. The results obtained for the BCA stock, RBF Kernel gives the model with the largest R2 error, which is around 0.978 - 0.982 and the Polynomial Kernel gives a prediction with the greatest accuracy, which is 98.14% - 99.14%. For BRI shares, RBF Kernel provides the model with the largest R2 error and prediction accuracy, which is 0.864 - 0.877 and 98.26% - 99.19%. It was found that the best Kernel function is RBF Kernel and the best data frequency is daily. Also obtained when the value of ?>, then the accuracy of the model and predictions will decrease. When the value ? <, the accuracy of the model increases but the accuracy of the prediction decreases. When the value of ? does not match (< or >), the accuracy of the model and prediction can increase or decrease, depending on the form of data. When the value of C<, the accuracy of the model and prediction decreases. When the value of C>, the accuracy of the model increases but the accuracy of the prediction decreases. It was found that the greater the ratio of training data from the test data, the accuracy of the model and its predictions will increase.