مشخصات پژوهش

This study applies educational data mining to predict and analyze student academic performance using a real-world dataset of 1000 students with demographic, behavioral and academic attributes. A preprocessing pipeline (data cleaning, missing value handling, categorical encoding, min–max normalization and train–test splitting) is followed by 3 analyses: supervised classification, K-means clustering and Apriori-based association rule mining. Decision Tree, Random Forest and Support Vector Machine classifiers are trained to predict a pass/fail label; the Support Vector Machine performs best, with accuracy = 0.69, F1 = 0.82, and recall = 1.00 for the pass class. Clustering reveals 7 distinct student profiles differing in study effort, prior achievement, attendance and parental support, while association rules show that high study hours and strong previous grades are linked to passing, and low study hours, low previous grades and low parental support to failing. The findings highlight study time, prior performance, attendance and family support as key determinants of academic success and support data-driven early intervention and personalized educational support.