Predicting Sleep Hours Based on Digital Device Usage

Project Overview

This project implements Linear Regression from scratch using Gradient Descent to predict the number of sleep hours based on a person’s age, daily screen time, device usage before sleep, and stress level. The dataset captures the impact of digital device usage on sleep quality and mental health.

---

Features

• Age
• Daily Screen Time (hours)
• Use of Devices Before Sleep (Yes/No)
• Stress Level

Target Variable:

• Sleep Hours

---

Project Highlights

• Manual implementation of Linear Regression (no sklearn LinearRegression used).
• Gradient Descent algorithm used to learn model parameters.
• Feature scaling using StandardScaler for efficient training.
• Visualization of training loss (MSE) and predicted vs actual sleep hours.
• Prediction for new individuals based on input features.

---

Results

• R² Score (Test Set): 0.61
• Mean Squared Error (MSE): 1.399
• Root Mean Squared Error (RMSE): 1.183 hours

---

Visualizations

1. Training Loss Curve (MSE vs Epochs)
2. Predicted vs Actual Sleep Hours

---

Tools & Libraries

• Python 3
• NumPy
• Pandas
• Matplotlib
• scikit-learn (for preprocessing and train-test split only)

---