Daily stock prices prediction using Bivariate Variance Gamma model

The Geometric Brownian Motion (GBM) model assumes that asset log returns are normally distributed. In fact, many asset log returns are found that are not normally distributed, therefore another model is needed, one of them is the Variance Gamma (VG) model. The VG process is a stochastic process. This process has three parameters which are generalizations of the Brownian motion that developed the Brownian. The added parameters are the drift of Brownian motion and the volatility of time changes. The VG process is used for modeling asset log returnsAs well as the GBM model, the VG model is widely applied in stock price modeling. Theoretically, the movement of the stock price of one company will be followed by the movement of the stock price of another company. Therefore, it would be more appropriate if the modeling was carried out together, namely using the Multivariate Variance Gamma (MVG) model. In this study, the Bivariate Variance Gamma (BVG) model was used. The data used is the daily closing price of stock of PT Bank Rakyat Indonesia (Persero) Tbk and PT Bank Negara Indonesia (Persero) Tbk. The parameter estimation was carried out using the maximum likelihood approach. Meanwhile, stock price prediction was carried out by using a Monte Carlo simulation approach. In this study, the prediction results using BVG obtained MAPE value less than10% which concludes that forecasting accuracy is excellent. © 2023 Author(s).