michaelozon/student-performance-factors-analysis-michael-ozon

🎓 Student Performance Factors — EDA & Insights

Michael Ozon — Assignment #1 (EDA & Dataset) Reichman University – Data Science Course

🎥 Presentation Video

https://drive.google.com/drive/folders/1cAXLzcZflMgv12EDlVTeQoKxzVumOjbd?usp=drive_link

📌 Project Overview

This project explores the Student Performance Factors dataset, containing 6,607 student records and 20 academic, behavioral, lifestyle, and demographic features.

The goal of this Exploratory Data Analysis (EDA) is to understand which factors most strongly influence students’ final exam scores (Exam_Score) and to extract clear, data-driven insights.

In this project I performed:

Data loading & validation

Data cleaning (missing values, duplicates)

Outlier inspection

Descriptive statistics

Correlation analysis

Multiple visualizations

Clear conclusions and interpretation

📊 Dataset Description

Source: Kaggle Rows: 6,607 Columns: 20 Target Variable: Exam_Score (continuous)

Feature Categories

Study-related: Hours_Studied, Attendance, Previous_Scores

Lifestyle: Sleep_Hours, Physical_Activity

Family: Family_Income, Parental_Education_Level

School environment: Teacher_Quality, School_Type

Personal: Gender, Motivation_Level, Internet_Access

Dataset Info Missing Values 🧼 Part 1 — Data Cleaning ✔ Missing Values

Three columns contained a small number of missing values:

Column Missing Teacher_Quality 78 Parental_Education_Level 90 Distance_from_Home 67

Action: Filled using mode() (most frequent category). Reason: Missingness was < 1.5% and all three variables are categorical.

✔ Duplicate Rows df.duplicated().sum()

Result: No duplicate rows found.

✔ Outlier Inspection

Outliers were inspected using boxplots. Some extreme values appeared in:

Hours_Studied

Tutoring_Sessions

Physical_Activity

However, all values were reasonable and consistent with realistic variation. No rows were removed.

📈 Part 2 — EDA (Exploratory Data Analysis) 📌 Descriptive Statistics 1️⃣ Attendance → Exam Performance

Insight: Attendance has the strongest connection to exam performance. Students with higher attendance consistently achieve higher scores. Correlation: 0.58

2️⃣ Hours Studied → Exam Performance

Insight: Clear positive relationship — more study hours lead to higher exam scores. Correlation: 0.45

3️⃣ Previous Scores → Exam Performance

Insight: Weak positive relationship. Previous academic performance contributes slightly but does not strongly predict the final exam score. Correlation: 0.18

4️⃣ Tutoring Sessions

Insight: Students with more tutoring sessions show slightly higher scores, but the effect is small. Correlation: 0.16

5️⃣ Gender Differences

Insight: Performance is nearly identical for males and females. Correlation: ~0

6️⃣ Sleep Hours

Insight: No meaningful relationship between sleep duration and exam score. Correlation: −0.017

7️⃣ Physical Activity

Insight: No significant relationship. Correlation: 0.03

8️⃣ Family Income

Insight: Students from Medium and High-income families have slightly higher median scores, but the effect is small and distributions heavily overlap.

🔥 Correlation Heatmap (Numerical Features) Key Predictors

Attendance — 0.58

Hours_Studied — 0.45

Weak Predictors

Previous_Scores (0.18)

Tutoring (0.16)

No Influence

Gender

Sleep_Hours

Physical_Activity

🧠 Final Insights & Conclusions 🎯 Strongest Factors Influencing Exam Performance

Class Attendance

Hours Studied

These two behavioral variables directly linked to study habits have the strongest predictive power.

🎯 Weak or No Influence

Sleep hours

Physical activity

Previous scores

Gender

Tutoring (minor effect)

Family income (minor effect)

🧩 Interpretation

The analysis suggests that academic engagement — especially consistent attendance and dedicated study time — is the primary driver of student success.

Lifestyle, demographic, and background factors play a much smaller role compared to active learning behaviors.