🎓 Student Performance Analysis & Prediction

GDGoC COMSATS WAH — DECODE: Data Science Bootcamp 2026

---

📌 Project Overview

This project applies a complete Data Science pipeline on a messy student dataset to analyze and predict final exam scores using Linear Regression. The dataset contains 205 rows and 14 columns with real-world data quality issues that were cleaned and processed before modeling.

---

📁 Dataset

Column	Type	Description
Student_ID	Numerical	Unique ID — dropped before modeling
Name	Text	Student name — casing issues fixed
Age	Numerical	Student age — wrong dtype fixed
Gender	Categorical	Male/Female — casing standardized
City	Categorical	Student city
Department	Categorical	CS / EE / ME / BBA / SE
Education_Level	Categorical	Intermediate / Bachelors / Masters
Attendance_%	Numerical	Attendance percentage
Study_Hours_Daily	Numerical	Daily study hours
Assignments	Numerical	Assignment score out of 10
Quizzes	Numerical	Quiz score out of 20
Midterm	Numerical	Midterm score
Internet_Access	Categorical	Yes / No
Final_Score	Numerical	TARGET — what we predict

---

🔧 Pipeline Steps

Step 1 — Data Loading

• Loaded dataset using pd.read_excel()
• Checked shape, dtypes, missing values, head/tail, describe

Step 2 — Data Cleaning

Fixed all 7 problems in the dataset:

• ✅ Gender casing → .str.title()
• ✅ Name casing → .str.title()
• ✅ Department inconsistencies → .str.upper().str.strip()
• ✅ Age wrong dtype → pd.to_numeric(errors='coerce') + median fill
• ✅ Impossible Attendance values → replaced with NaN
• ✅ Impossible Final_Score values → replaced with NaN
• ✅ Missing values → filled with median (robust to outliers)
• ✅ Duplicate rows → removed with drop_duplicates()

Step 3 — EDA + Visualization

7 charts with written interpretations:

1. Distribution of Final_Score (histplot)
2. Attendance % vs Final_Score (scatter)
3. Study Hours vs Final_Score (scatter)
4. Midterm vs Final_Score (scatter)
5. Correlation Heatmap
6. Final_Score by Department (boxplot)
7. Students per Department (countplot)

Step 4 — Feature Engineering + Pipeline

New features created:

• Total_Academic = Midterm + Assignments×5 + Quizzes×2
• Attendance_Category = Low / Medium / High (binned)
• Study_Efficiency = Study_Hours × (Attendance / 100)

Encoding decisions:

• OneHot → Gender, Internet_Access, City, Department (no order)
• Ordinal → Education_Level, Attendance_Category (order exists)
• StandardScaler → all numerical columns

Step 5 — Model + Evaluation

• Model: Linear Regression via sklearn Pipeline
• Train/Test Split: 80/20, random_state=42

---

📊 Results

Metric	Value
MAE	~2.45
RMSE	~3.27
R²	~0.59

The model explains 59% of the variation in Final_Score. The strongest predictor was Midterm score, followed by the engineered Total_Academic feature.

---

🚀 What I Would Improve

1. Try Random Forest / XGBoost to capture non-linear patterns
2. Collect more data — 200 rows is small for ML
3. Add stronger features like previous GPA or assignment submission rate

---

🛠️ Tech Stack

• Python 3.10
• Pandas, NumPy
• Scikit-learn (Pipeline, ColumnTransformer, LinearRegression)
• Matplotlib, Seaborn
• Google Colab

---

📂 How to Run

1. Clone this repo

git clone https://github.com/your-username/student-performance-prediction

2. Upload notebook to Google Colab

3. Upload dataset to Google Drive at:

/content/drive/MyDrive/gdg project/student_performance_dataset.xlsx

4. Run all cells top to bottom

---

👤 Author

Muhammad Majid Ali COMSATS University Islamabad, Wah Campus GDGoC DECODE Data Science Bootcamp 2026