My Approach:

This project explores how machine learning and data visualization can identify behavioral patterns linked to social-media overuse.
Using a dataset of 1,500+ survey responses, I trained multiple classification models to predict whether a user is likely to be addicted to social media, based on psychological, demographic, and activity-based indicators.
To make the insights accessible, I built an interactive Power BI dashboard that visualizes trends in screen-time, emotional dependency, and platform usage habits.

Vision and Innovation

To develop a predictive system that quantifies a person’s risk of social-media addiction and communicates insights clearly through an analytical dashboard. helping educators, parents and researchers understand key behavioral triggers.

Machine Learning Process

1. Data Pre-Processing: Cleaned and normalized survey data; handled missing values and categorical encoding.

2. Feature Selection: Identified high-impact variables such as hours spent online, sleep quality, engagement frequency, and mood correlation.

3. Model Development: Decision Tree classifier achieved 78 % accuracy. Benchmarked against Logistic Regression and Random Forest models using k-fold cross-validation.

4. Model Optimization: Performed pruning and feature-importance analysis to improve generalization and reduce overfitting.

Tools used

1. Languages: Python (pandas, scikit-learn, matplotlib)

2. Visualization: Power BI (DAX, Python integration)

3. Data Handling: CSV, NumPy

4. Version Control: GitHub

Key findings

1. Users spending more than 5 hours/day on social media showed a 60 % higher addiction-risk score.

2. Late-night usage and irregular sleep cycles were the strongest predictors of addiction.

3. Female users aged 18–25 showed higher engagement consistency across multiple platforms.