📱 Smartphone Stress Detection System

A web app that predicts your stress level based on your daily smartphone usage habits — and tells you whether your stress is getting better, worse, or staying the same over time.

---

🤔 What Is This?

Most stress apps just tell you "you're stressed right now." This one goes further — it tracks your stress over time and predicts the direction it's heading. Are things getting worse? Are you recovering? It'll tell you, and it'll alert you before things get critical.

Built as part of a pervasive computing project, this app uses a Random Forest model trained on real mobile usage data to predict stress levels from your daily phone habits like screen time, sleep, social media usage, and more.

---

✨ Features

1. 🔮 Stress Trajectory Prediction

Not just "your stress is HIGH" — but "your stress has been increasing for 3 days, take action now." Uses a 7-day sliding window and linear regression to detect trends.

2. 🧠 Explainable Predictions

Tells you why the model made that prediction — which factors contributed the most (e.g., "High screen time and low sleep are your biggest stress drivers").

3. 🔒 Consent-Aware Privacy

You choose how much data to share:

• Full consent → All 12 features used, detailed results
• Limited consent → Only broad features (screen time, sleep, activity)
• Minimal consent → Just screen time and sleep

4. 🚨 Intelligent Alert System

Smart alerts that don't spam you. Alerts adapt based on:

• Your consent level (more detail = more specific alert)
• Time of day (night-time high stress triggers a "digital detox" alert)
• Repeated stress spikes (escalates to critical alert after 3+ high readings)
• 30-minute cooldown between same-type alerts

5. 📄 PDF Report Download

Download a full report of your prediction including trajectory, explanation, contributing factors, and consent level used.

---

🛠️ Tech Stack

Layer	Technology
Backend	Python, Flask
ML Models	Random Forest (Classifier + Regressor)
Data Balancing	SMOTE (imbalanced-learn)
Trend Analysis	Linear Regression (scipy)
Visualizations	Matplotlib, Seaborn
PDF Generation	ReportLab
Frontend Storage	Browser localStorage (for user history)

---

📂 Project Structure

stress-detection-app/
│
├── app.py                          # Main Flask app + all ML logic
├── stress_model.pkl                # Pre-trained model
├── requirements.txt                # Python dependencies
│
├── static/
│   ├── confusion_matrix.png        # Auto-generated on startup
│   └── feature_importance.png      # Auto-generated on startup
│
├── templates/
│   └── index.html                  # Frontend UI
│
├── STRESS_TRAJECTORY_EXPLANATION.md
├── INTELLIGENT_ALERT_SYSTEM.md
├── HOW_TO_GET_3_HISTORICAL_READINGS.md
└── RUN_LOCALHOST.md

---

🚀 How to Run Locally

Step 1: Clone the repo

git clone https://github.com/SUBASHREE15S/Stress-Detection.git
cd Stress-Detection

Step 2: Install dependencies

pip install -r requirements.txt

Step 3: Add your dataset

Place your mobile_addiction_data.csv file somewhere on your system and update this line in app.py:

csv_path = r"C:\path\to\your\mobile_addiction_data.csv"

Step 4: Run the app

python app.py

Step 5: Open in browser

http://127.0.0.1:5000/

---

📊 Input Features

The model uses these 12 features from your daily phone usage:

Feature	Description
Daily Screen Time (hrs)	Total hours phone screen was on
Phone Unlocks Per Day	How many times you picked up your phone
Social Media Usage (hrs)	Time on Instagram, Twitter, etc.
Gaming Usage (hrs)	Time spent on mobile games
Streaming Usage (hrs)	YouTube, Netflix, etc.
Messaging Usage (hrs)	WhatsApp, SMS, etc.
Work-Related Usage (hrs)	Emails, work apps
Sleep Hours	Hours of sleep per night
Physical Activity (hrs)	Exercise or movement
Time With Family (hrs)	Offline social time
Online Shopping (hrs)	Time on shopping apps
Push Notifications/Day	Number of notifications received

---

🎯 How Stress Is Calculated

Stress Score =
  (Screen Time × 2) + (Unlocks × 0.3) + (Social Media × 1.5)
  + (Gaming × 1.5) + (Streaming × 1.2) + (Work Usage × 1.3)
  + (Notifications × 0.2)
  − (Sleep × 4) − (Physical Activity × 3) − (Family Time × 2)

The score is then classified into Low / Medium / High using the 33rd and 66th percentiles.

---

📈 Trajectory Prediction — How It Works

1. Every prediction you make is saved with a timestamp
2. After 3 or more predictions, linear regression calculates the trend
3. Based on the slope:
   • Slope > +2.0 → INCREASING ⚠️
   • Slope < −2.0 → RECOVERING ✅
   • In between → STABLE 📊

To quickly test this, click the "🧪 Add Test History" button in the app — it loads 3 sample readings instantly.

---

🚨 Alert Types

Alert	Trigger	Severity
Immediate	High stress + increasing trend	🟠 High
Early Warning	Moderate stress + rising trend	🟢 Low
Digital Detox	High stress during night hours (10PM–6AM)	🔵 Medium
Escalated	3+ consecutive high stress readings	🔴 Critical

Alerts have a 30-minute cooldown — no spam. Critical alerts never auto-dismiss.

---

📋 Requirements

Flask==3.0.0
pandas==2.1.4
numpy==1.26.2
scikit-learn==1.3.2
imbalanced-learn==0.11.0
matplotlib==3.8.2
seaborn==0.13.0
reportlab==4.0.7
scipy==1.11.4

---

🔮 Future Improvements

• Connect to real phone usage APIs (Android Digital Wellbeing, iOS Screen Time)
• Push notifications to mobile
• Persistent user accounts across devices
• Integration with wearables for heart rate and sleep data

---

👩‍💻 Built By

ANU SWATHI V.R

SUBASHREE S

LAHARI K.R