Comparison of stock price prediction using geometric Brownian motion and multilayer perceptron
The Stock is defined as an investor ownership, sign of their investment, or the amount of fund invested in a company. In the transaction process of stock exchange, stock is the most traded instrument. Thus, the forecasting of stock price is very important to develop an effective market trading strat...
Gespeichert in:
Hauptverfasser: | , , |
---|---|
Format: | Tagungsbericht |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The Stock is defined as an investor ownership, sign of their investment, or the amount of fund invested in a company. In the transaction process of stock exchange, stock is the most traded instrument. Thus, the forecasting of stock price is very important to develop an effective market trading strategy. The forecasting of stock prices can anticipate investment losses and provide optimal benefits for investors. In this paper, Microsoft stock prices will be predicted by the geometric Brownian motion and multilayer perceptron methods. Prediction of stock prices using geometric Brownian motion was begun by calculating the return value of the data. Then, the normality test of the return value is carried out. The value of return must be normally distributed. Then, we did calculations to get the value of drift and volatility. The parameters of geometric Brownian motion are assumed constant. The parameter values will be used as input in prediction process with MATLAB. In the multilayer perceptron method prediction, the data divided into two parts, 70 % for training data and 30 % for validation data. Then, the data must be normalized first. In the prediction process using multilayer perceptron, we will initialize the weight, correct the weight values, correct bias and calculate error value. As the result, the MAPE value from multilayer perceptron method predictions was 0.05266. This value obtained when the number of neurons are 2 in the input and the number of neurons are 3 in the hidden layer. Meanwhile, the MAPE value produced by geometric Brownian motion method were 0.0221 when 10 trajectories and 0.019571689 when 10000 trajectories. Then, the result of geometric Brownian motion method is better than multilayer perceptron method. |
---|---|
ISSN: | 0094-243X 1551-7616 |
DOI: | 10.1063/5.0008066 |