Sabanci University DSA210 Introduction to Data Science Course Fall 2024-2025 Term Project.
This project analyzes the relationship between screen time and notifications received across different apps.
For the final report, see here.
The motivation for this project stems from a curiosity about how app notifications impact screen time. By exploring this relationship, I aim to uncover patterns in my own device usage and derive actionable insights.
- Jupyter Notebook: For coding and documentation.
- Pandas: For data manipulation and analysis.
- Matplotlib and Seaborn: For creating static visualizations.
- Scikit-learn: For linear regression modeling.
- Excel: For initial data collection and storage.
The data for this project was manually collected from my personal device over a period of several days. It includes the following metrics:
- Screen time (minutes): Time spent on each app per day.
- Notifications: Number of notifications received from each app per day.
- Pickups: Number of times the device was accessed.
The dataset can be found here.
The collected data underwent cleaning and preprocessing in main.ipynb. Key steps included:
- Removing incomplete entries.
- Aggregating data to calculate daily totals for screen time and notifications.
- Generating derived metrics like screen time per notification.
Visualizations were created to explore trends and relationships within the dataset. Highlights include:
- Scatter plots to visualize app usage trends day by day.
- Bar charts for total screen time and notifications per day.
- Dual-axis plots to compare screen time and notifications.
The primary analysis involved:
- Investigating the correlation between screen time and notifications.
- Performing linear regression to quantify the relationship between the two variables.
Findings and detailed analysis steps can be found in main.ipynb.
- There is no significant correlation between screen time and notifications.
- Linear regression shows a weak relationship, with a very low R² value.
For detailed findings, see the final report.
- Data Scope: The dataset is limited to personal usage over a short time span, making generalizations difficult.
- Data Completeness: Not all apps reported notifications accurately, leading to potential biases.
- Expand the Dataset: Collect data over a longer period and from multiple devices.
- Integrate Additional Metrics: Include metrics like app categories or usage goals.
- Advanced Analysis: Explore machine learning techniques for deeper insights.