การพยากรณ์มูลค่าทรัพย์สินสุทธิกองทุนรวมในกลุ่มธุรกิจการดูแลสุขภาพโดยใช้การเรียนรู้ของเครื่อง

Abstract

Investing in mutual funds is one of the most popular financial products for individuals seeking long-term financial planning and establishing a retirement fund. This is because it is a suitable investment option for the long term and provides relatively favorable returns. The recent COVID-19 pandemic has significantly impacted the economic downturn in Thailand, affecting various industries and businesses. However, the healthcare sector has continued to grow steadily both before and after the COVID situation. This growth is largely attributed to changes in consumer behavior and the aging population, making investments in securities, particularly mutual funds in the healthcare sector, highly appealing. Therefore, studying price forecasting can be a useful tool for decision-making, evaluating, and managing investment risks effectively. However, it is important to note that previous studies on security price forecasting often rely solely on historical trading data for future predictions, disregarding other influencing factors affecting price fluctuations. This study, therefore, proposes a study on forecasting mutual fund prices in the industrial and healthcare sectors by incorporating various influential factors through a literature review. The study utilizes Deep Learning (DL) models, specifically Artificial Neural Networks and Long-Short Term Memory (LSTM), Machine Learning (ML) models, including Support Vector Regression, and statistical time series models, including Seasonal Autoregressive Integrated Moving Average with exogenous variables (SARIMAX). The study focuses on six mutual funds in Thailand. The findings reveal that incorporating various influential factors significantly enhances the accuracy of price forecasting. Statistical models demonstrate comparable forecasting efficiency to DL models and outperform ML models. Furthermore, the research illustrates the practical application of these models for risk assessment and management, particularly in conjunction with the Vector Autoregressive (VAR) time series model. The results indicate a consistent trend between predicted and actual prices.