Machine Learning — Assignment 1 (AITU)

Classification of student academic performance using Decision Tree, Random Forest, and K-Nearest Neighbours on a real student performance dataset (1,000,000 records).

---

Overview

Dataset: student_performance.csv — 1,000,000 student records with four behavioural/academic features and a five-class grade label (A / B / C / D / F).

Feature	Description
weekly_self_study_hours	Hours spent studying outside class
attendance_percentage	Percentage of classes attended
class_participation	Participation score
total_score	Composite academic score
grade	Target label: A / B / C / D / F

Goal: Build and compare multi-class classifiers that predict a student's grade category from these features.

---

Notebook structure

Assignment1_Adil_Ormanov.ipynb — fully executed notebook with all cell outputs saved:

Section	Content
1. Imports	Libraries, global settings
2. Load Data	Read full CSV; draw stratified 100k sample
3. EDA	Descriptive stats, missing-value check, histograms, correlation heatmap, box plots by grade
4. Preprocessing	LabelEncoder for grade target; StandardScaler; 80/20 stratified split
5. Model Training	DecisionTreeClassifier, RandomForestClassifier (100 trees), KNeighborsClassifier (k=5)
6. Evaluation	classification_report per model; confusion-matrix plots; grouped bar chart comparison
7. Split Sensitivity	Metrics vs test-size ratio (20 %–40 %), line plots per model
8. Feature Importance	Horizontal bar charts for Decision Tree and Random Forest
9. Export	Results saved to student_performance_results.csv

---

Results (80/20 test split — 20,000 test samples)

Model	Accuracy	Precision	Recall	F1-Score
Decision Tree	0.9979	0.9978	0.9979	0.9978
Random Forest	0.9981	0.9982	0.9981	0.9981
KNN (k=5)	0.9705	0.9705	0.9705	0.9704

All weighted averages. Decision Tree and Random Forest are near-identical in performance; KNN trails by ~2.7 pp on F1. Random Forest is the strongest single model (F1 = 0.9981).

---

Files

File	Description
Assignment1_Adil_Ormanov.ipynb	Fully executed notebook (outputs included)
student_performance.csv	Input dataset (1,000,000 rows)
student_performance_results.csv	Per-model metrics across 5 split experiments
requirements.txt	Python dependencies

---

Getting started

# 1. Create and activate a virtual environment
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate

# 2. Install dependencies
pip install -r requirements.txt

# 3. Open the notebook
jupyter lab Assignment1_Adil_Ormanov.ipynb

---

Adil Ormanov — GitHub