Accurate stock forecasting in the Thai capital market remains challenging due to high volatility and pronounced non -linear dynamics in financial time-series. This study conducts a comparative evaluation of machine learning (ML) models for one-day-ahead prediction of daily closing prices of stocks listed on the Stock Exchange of Thailand (SET). Two widely used technical indicators-Simple Moving Average (SMA) and Exponential Moving Average (EMA)-are employed as input features to capture trend and momentum. Five years of daily data (2019–2023) for three large-cap stocks (PTT. BK, BBL.BK, and CP ALL.BK) are used to assess model robustness under multiple market conditions. We examine three train-test splits (70:30, 80:20, 90: 10) to evaluate stability across data regimes. Experiments show that Linear Regression consistently outperforms competing approaches, achieving the highest coefficient of determination (R2) of 0.9758, and producing one-day-ahead forecasts closely aligned with actual closes for PTT.BK, BBL.BK, and CPALL.BK. The findings indicate that a systematic pipeline grounded in carefully selected technical indicators can provide an effective framework for stock-trend analysis and support investment decision-making in the volatile Thai equity market.

This research paper examines the effectiveness of different machine learning (ML) models in analyzing how library usage influences the academic performance of Thammasat University students. By leveraging a comprehensive dataset that contains library interactions such as library of things (LoT) services, book borrowing and returning, and space services, this study evaluates five ML models: decision tree, k-nearest neighbors, Naïve Bayes, random forest, and support vector machine. The primary goal is to identify which library services and facilities significantly impact student success to provide actionable insights for customizing library offerings and enhancing academic outcomes. Results were that the decision tree model was particularly effective, delivering precise, reliable classification of factors affecting student academic performance into training data and data testing, using the 10-fold cross-validation procedure. These findings help demonstrate practical applications of ML in educational contexts and facilitate library strategic planning and management.