Student Academic Performance Analysis

395 Students | 33 Variables | 5 Research Questions

Exploring factors associated with final mathematics grades among secondary school students in Portugal.

Source: UCI Machine Learning Repository — Student Performance Dataset

Website: UCI Student Performance Dataset

Research: Cortez and Silva (2008)

Dataset: student-mat.csv

• 395 students from two secondary schools in Portugal
• 33 original variables covering academic, personal, and family characteristics
• Target variable: Final mathematics grade (G3), scored from 0 to 20
• Key factors: Study time, school absences, previous failures, and earlier grades

Research goal: Identify factors associated with final mathematics performance.

Data Quality

• 395 students and 33 original variables
• 0 missing values and 0 duplicate records
• No observations removed during initial cleaning

Variable Transformations

• Study time: Four categories ranging from less than 2 hours to more than 10 hours per week
• Final grade classification: Pass (G3 ≥ 10) or Fail (G3 below 10)

Key Result

265 students passed (67.1%) | 130 failed (32.9%)

Average grade: 10.42 / 20 | Median: 11

Key finding: Most grades fall between 8 and 15, with 38 students (9.6%) receiving zero.

Highest average: 5–10 hours per week (11.40 / 20)

Lowest average: Less than 2 hours per week (10.05 / 20)

Key finding: Greater study time is associated with somewhat higher average grades, but the relationship is not strictly increasing.

Pearson correlation: r = 0.0342

Sensitivity analysis: r = 0.0845 for students with 20 or fewer absences

Key finding: School absences showed almost no linear association with final grades.

No previous failures: Average final grade = 11.25

Three or more failures: Average final grade = 5.69

Key finding: More previous academic failures were associated with substantially lower final mathematics grades.

First-period grades (G1) vs. final grades (G3): r = 0.8015

Second-period grades (G2) vs. final grades (G3): r = 0.9049

Key finding: Earlier mathematics grades, especially G2, showed strong positive associations with final performance.

MAJOR FINDINGS

• Earlier grades: G2 had the strongest correlation with G3 (r = 0.9049).
• Previous failures: More failures were associated with lower grades.
• Study time: The 5–10 hours/week group averaged 11.40.
• Absences: Negligible linear correlation with G3 (r = 0.0342).

STUDY LIMITATIONS

• Data from two Portuguese schools
• 38 students received zero final grades for unknown reasons
• Correlations do not establish causation

MAIN TAKEAWAY: Earlier academic performance, particularly second-period grades, was the strongest correlate of final mathematics achievement among the variables examined.

Student Academic Performance Analysis

Bashirul Shikder | Arizona State University | DAT 301

Special Acknowledgment

I would like to express my sincere gratitude to Dr Asiyeh Rafieipour, my DAT 301 instructor, for her guidance, instruction, and support throughout the course.

Thank you for your time and attention!

Cortez, P. (2008). Student performance [Data set]. UCI Machine Learning Repository. https://doi.org/10.24432/C5TG7T

Cortez, P., & Silva, A. (2008). Using data mining to predict secondary school student performance. In Proceedings of the 5th Future Business Technology Conference (FUBUTEC 2008) (pp. 5–12).

UCI Student Performance Dataset