PREDIKSI INDEKS HARGA SAHAM GABUNGAN MENGGUNAKAN SUPPORT VECTOR REGRESSION DENGAN ALGORITMA GRID SEARCH

Abstract

The existence of capital market Indonesia is one of the important factors in the development of the national economy, proved to have many industries and companies that use these institutions as a medium to absorb investment and media to strengthen its financial position. Capital market Indonesia is an emerging market development is very vulnerable to global economic conditions and capital markets of the world. Prediction JCI (Jakarta Composite Index) is necessary to know the great value that will occur in the future so as investors can take the right policy. To predict in this study used a Support Vector Regression (SVR) method to find the hyperplane in the best regression function to predict the closing price of the JCI using a linear kernel function with output in the form of continuous data. Parameter selection cost and epsilon using a grid search algorithm combined with cross validation and obtained best cost 1 and best epsilon 0.1. While the criteria to measure the goodness of the model is MAPE (Mean Absolute Percentage Error) and R2 (Coefficient Determination). The results of this study showed that SVR with linear kernel function provides excellent accuracy in the prediction of JCI with R2 results on training data 98.4% with a MAPE 0.873% while the testing of data R2 90.9% with a MAPE 0.613%. Keywords: JCI, Support Vector Regression (SVR), Hyperplane, Kernel Linear, Grid Search Algorithm, Cross Validation, Accuracy