Overview

This project builds a production-ready machine learning pipeline to predict smartphone addiction using behavioral data such as screen time, app usage, sleep patterns, and user activity.

It goes beyond basic modeling by implementing data preprocessing, feature engineering, class balancing, model training, and a Streamlit-based frontend.

Objective

• Predict whether a user is addicted to smartphone usage
• Analyze behavioral patterns affecting addiction
• Deploy a real-time prediction system

Dataset Features

• Demographics: Age, Gender
• Usage Behavior: Screen time, Social media, Gaming, Work/Study
• Activity Metrics: Notifications, App opens
• Lifestyle: Sleep hours, Stress level
• Target: addicted_label (0 = Not Addicted, 1 = Addicted)

ML Pipeline

1. Data Preprocessing

• Removed irrelevant features (IDs)
• Handled missing values
• Encoded categorical variables

2. Feature Engineering

• Total Usage Time
• Sleep Deficit

3. Handling Imbalance

• Applied SMOTE to balance classes

4. Model

• Random Forest Classifier

5. Evaluation

• Accuracy
• Classification Report
• Confusion Matrix

Results

The model effectively captures behavioral patterns and predicts addiction with strong performance after balancing the dataset.

Frontend

Built with Streamlit:

• User-friendly interface
• Real-time predictions
• Interactive input controls

Tech Stack

• Python
• Scikit-learn
• Pandas / NumPy
• Matplotlib
• Imbalanced-learn (SMOTE)
• Streamlit

Run Locally

pip install -r requirements.txt
python smartphone_addiction_pipeline.py
streamlit run app.py