Prediksi Harga Saham Menggunakan Geometric Brownian Motion Termodifikasi Kalman Filter dengan Konstrain

Abstract

An attractive profit is one of the attractions offered by stock investment. Changes in stock prices that are difficult to predict will result in uncertain value of profits, so it is necessary to predict the stock price using forecasting method. The model used is Geometric Brownian Motion (GBM). This model can predict future stock price movements based oh historical stock data. Forecasting results with the Geometric Brownian Motion model produce significant errors due to constant parameters. To reduce the values of error, it is necessary to add a filtering method that is Kalman Filter (KF) by limiting the state variables using norm. Historical data was taken from 3 different closing price stock data, namely shares of Bank BRI, PT. Telekomunikasi Indonesia Tbk, and Unilever Indonesia with period of January 1 – December 31, 2019. Based on the results obtained, the addition of contraints on the GBM-KF model does not significantly influence the MAPE value. At the forecasting stage using testing data with GBM-KF model without constraints, the average MAPE  value for BBRI was 0.1122%, TLKM 0.0899%, and UNVR 0.0678%. While forcasting using GBM-KF model with constrains, the average MAPE value for BBRI was 0.0958%, TLKM 0.0808%, and UNVR 0.0674%. The values of MAPE obtained are included in the high accuracy forecasting category.