IMPLEMENTATION GRID SEARCH OF RBF AND POLYNOMIAL ON SUPPORT VECTOR REGRESSON FOR CLOSING STOCK PRICES PREDICTION ON PT INDOFARMA (INAF)

Abstract

Stocks represent evidence of ownership of an asset. The highly volatile nature of stock prices makes it difficult for investors to predict stock prices, necessitating the analysis of stock investments. This research aims to forecast for the next 30 days the closing price of PT Indofarma (INAF) stocks using the best model, and the accuracy level of the employed model was analyzed based on the data from the last seven years. The research used the Support Vector Regression (SVR) method, which is known for its capability to handle nonlinear data through kernel functions. The Radial Basis Function (RBF) and polynomial kernels are used in this case. The challenge with SVR lies in determining the optimal hyperparameter, which can be addressed through hyperparameter tuning using grid search. The research results show that the best model is the SVR kernel RBF model with optimal hyperparameter C=1,γ=0.01, and ε=0.01. Based on the performance evaluation results of the best model, the MAPE, MSE, and MAE values are equal to 1.537%,1483.936, and 23.409.