🧠 Student Depression Prediction Using Machine Learning

📖 Project Overview

Depression among students is a serious concern affecting academic performance and overall well-being.
This project analyzes student data, visualizes patterns using EDA, and builds predictive models to identify students at risk. Models such as Logistic Regression, Random Forest, SVM, Decision Tree, KNN, Naive Bayes, Gradient Boosting, AdaBoost, Bagging, and XGBoost are evaluated to find the most effective approach.

Kaggle Notebook: Student Depression Prediction

🎯 Objectives

Analyze key factors contributing to student depression
Build predictive ML models to classify depression risk
Perform EDA to understand feature relationships
Evaluate model performance using accuracy, precision, recall, and F1-score
Identify the best-performing machine learning model for prediction

✨ Features

Exploratory Data Analysis (EDA) including:
- Sleep duration vs depression
- Gender vs depression
- Financial stress vs suicidal thoughts
- Work/study hours vs depression
- Academic pressure vs depression
- Pair plot of all features
Multiple ML models for classification
Performance metrics for each model: accuracy, precision, recall, F1-score

🛠 Tools & Technologies

Python
Pandas & NumPy (Data manipulation)
Matplotlib & Seaborn (Visualization)
Scikit-learn (Machine Learning models & evaluation)
XGBoost & AdaBoost (Advanced boosting techniques)
Jupyter Notebook (Development environment)

📂 Dataset

Source: Student Depression Dataset
Dataset Name: Student Depression Dataset
Number of Samples: 5,581
Target Variable: Depression (0 = Not Depressed, 1 = Depressed)
Features Include:
- Sleep duration
- Gender
- Financial stress
- Work/study hours
- Academic pressure
- Suicidal thoughts

📊 Model Performance

Model	Accuracy
Logistic Regression	0.8364
Random Forest	0.8319
Support Vector Machine	0.8344
Decision Tree	0.8034
K-Nearest Neighbors	0.8169
Naive Bayes	0.8212
Gradient Boosting	0.8375
AdaBoost	0.8389
Bagging Classifier	0.8233
XGBoost	0.8208

Best Accuracy Achieved: AdaBoost (83.89%)
Classification reports for all models show high precision, recall, and F1-scores for both depressed and non-depressed classes.