APPLICATION OF GAUSSIAN PROCESS AND SARIMAX METHODS FOR PREDICTING STOCK TREND MOVEMENTS

• Main Author: Bella Adhina, Rossa
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76227
• Institution: Institut Teknologi Bandung

Physics is the natural science that studies matter, motion, and behavior through space and time. The field of physics has evolved to encompass complex systems, involving entities capable of self-organization, with wide-ranging applications in technology, economics, mathematics, and other domains. The author experienced significant losses in investments and turned to physics to reduce the risk of future losses in investments. With the advancing times, especially in technology and economics, society becomes increasingly aware of future needs, prompting them to consider investments. Investment is an economic action to meet future needs by deferring expenditures in the present. In the financial market, stocks are the most favored investment instrument among the public. Leveraging knowledge gained during their time as a physics student, the author applies various concepts to minimize the risks when investing in stocks. The author employs the Gaussian Process and SARIMAX methods to forecast the movement of Aneka Tambang's (ANTM.JK) stock, using daily closed data from Yahoo Finance!. The results reveal SARIMAX is more accurate for short-term forecasts, whereas Gaussian Process performs better for long-term predictions. Gaussian Process requires longer computation time due to complex matrix calculations and large datasets. On the other hand, SARIMAX is more efficient with a simpler structure. Stock prices are often influenced by unpredictable external factors like global events, news, and market sentiment. Gaussian Process tends to focus on noise in historical data, which may lead to unstable or inaccurate price movement predictions. In contrast, SARIMAX models trends and recurring patterns, providing more stable predictions. SARIMAX's parameter interpretation is clearer, aiding in understanding the factors affecting stock price movements, while Gaussian Process parameters are more abstract and challenging to interpret.