Lazaro Martull

Circadian Rhythm Analysis Using Twitter Activity

Overview

This project analyzes global activity patterns using Twitter timestamps as a proxy for daily sleep–wake rhythms. The objective was to examine how geographic location (time zone) and occupation relate to shifts in online activity that may indicate circadian rhythm disruption.

Data

Dataset: jobs_sleepwalk (2020)
Scale: 4.5M+ records

Key fields:

• characteristic — occupation / role
• utc_timestamp — posting time (UTC)
• user_hash — anonymized user identifier
• location — user-reported location

Methodology

• Converted UTC timestamps into hourly and temporal features
• Cleaned and standardized location data
• Filtered to top locations and occupations for interpretability
• Visualized patterns using bar charts, hourly activity plots, and heatmaps

Key Results

• Analyzed 4.5M+ Twitter records to identify global circadian activity patterns
• Observed distinct regional posting cycles consistent with time-zone effects
• Identified occupation groups with elevated late-night activity behavior

Tools & Technologies

• Python
• Pandas
• Matplotlib / Seaborn / Plotly

Code & Analysis

The full data preprocessing, exploratory analysis, and visualization workflow was implemented in Python using Pandas and visualization libraries. The notebook includes timestamp transformations, location filtering, and multiple plots used in the final report.

Limitations

Twitter activity is a behavioral proxy and does not directly measure sleep duration or sleep quality.

Files

• 📄 Final Report: report.pdf
• 🧪 Analysis Notebook: Project_Circadian_Rhythms.ipynb