This study applies a first-order Markov chain to analyze and model the academic progression of 317 students from the Faculty of Education at Thu Dau Mot University, utilizing their semester Grade Point Averages (GPA) as the core data. Students' GPAs were methodologically classified into four distinct academic performance states: Weak (0–4.99), Average (5.0–6.99), Good (7.0–7.99), and Excellent (8.0–10.0). Transition matrices were constructed to capture the movements between these performance states across consecutive semesters. Descriptive analysis reveals a positive performance trend, specifically a frequent transition from the Average to the Good group, and a high level of stability observed within the Excellent group, particularly in the later stages of the program. A crucial Chi-square test for homogeneity revealed statistically significant differences, indicating that the learning process is non-homogeneous over time, reflecting fluctuations in student learning behavior. However, to fulfill the objective of forecasting the expected distribution of student performance in the subsequent semester, a weighted average transition matrix was computed, giving greater emphasis to the influence of more recent academic data. Forecasting results suggest that approximately 90% of students are expected to concentrate within the Good and Excellent categories, confirming a high standard of academic performance and providing valuable empirical evidence for targeted student support and curriculum management within the Faculty of Education.

Based on previous surveys, it is evident that Food Technology students at Thu Dau Mot University face challenges in learning English, such as limited vocabulary, weak grammar, inaccurate pronunciation, and poor listening and speaking skills. These can lead to declining academic performance and restricted career opportunities. The paper highlights the potential of Artificial Intelligence (AI) in overcoming these challenges. AI tools like Gemini Google can provide personalized learning experiences, improve interaction with the language, and offer immediate feedback. Gemini is specifically chosen due to its free access, Vietnamese interface, and support for over 40 languages. This study demonstrates that Gemini is an effective tool for undergraduate students, particularly those in Food Technology who have weak English backgrounds, to improve their English skills. Gemini improves vocabulary, grammar, speaking, listening, pronunciation, reading, and writing. It offers simple instructions and ideas in both English and Vietnamese, making it ideal for beginners, selflearners, and people with limited English proficiency. Unlike traditional learning, Gemini provides access at any time and from any location, as well as the ability to overcome shyness and blunders. Gemini's large, up-to-date database and user-friendly interface enable personalized learning paths and can be paired with other methods for best language learning. The study also proposes various strategies for utilizing Gemini Google to improve English learning effectiveness in areas like vocabulary, grammar, listening, speaking, reading, writing, and creating a positive learning environment. Furthermore, Gemini's value extends beyond languages, offering assistance in a variety of other areas.