Measuring the effect of certain basic, social and academic activities on success using modern technique

Academic success in education is shaped by the complex interaction of numerous personal, familial, environmental, and pedagogical variables. Factors such as students' study habits, sleep patterns, time allocated to leisure activities, interest in homework, and general attitudes toward learning have a direct impact on academic outcomes. In addition, students' levels of academic self-efficacy and their cognitive-emotional responses to challenging subjects—particularly mathematics—play a critical role in determining achievement. In recent years, there has been growing interest in analyzing these multidimensional factors, and the development of models to predict academic performance has become a significant area of research. In this context, artificial intelligence applications—especially fuzzy logic-based models—stand out as powerful tools for analyzing complex structures characterized by uncertainty and imprecise data. This thesis proposes a fuzzy logic-based model designed to predict students' written mathematics exam scores. The model incorporates fundamental academic and behavioral input parameters, including daily sleep duration, daily playtime, daily study time, daily homework time, and students' self-reported levels of mathematical hopelessness. A Sugeno-type fuzzy inference system and an Adaptive Neuro-Fuzzy Inference System (ANFIS) architecture were structured to transform vague and uncertain data into meaningful outcomes, using rule sets developed with expert consultation. Each variable in the model is represented at low, medium, and high levels using triangular membership functions. A total of 243 "If… then…" rules were defined, and the model was trained using a hybrid learning algorithm combining backpropagation and the least squares method. Unlike models tested with synthetic data, this model was evaluated using real data collected from a selected student group, ensuring realistic student profiles. The results demonstrate that the ANFIS-based approach can predict academic success with high accuracy and outperforms traditional methods in modeling complex relationships. By capturing the nuances of student behavior, this model offers valuable insights for early intervention and the development of personalized learning strategies. In conclusion, this thesis emphasizes the importance of integrating AI-assisted decision-making processes into the education system and provides new opportunities for teachers, parents, and educational administrators to monitor and enhance student performance. The use of the ANFIS-based model contributes to the field of educational data mining and highlights the potential of intelligent systems that support adaptive learning environments, aiming to offer a more inclusive, equitable, and high-quality educational experience for all students.