Global Shark Incident Analysis

A reproducible data-cleaning and exploratory-analysis project built from a historical shark-incident register. The analysis standardizes dates, locations, activities, sex, fatality status, and species text, then examines the composition of recorded incidents.

What changed from the bootcamp submission

The original team repository contained duplicate notebooks, drafts, hard-coded Windows paths, and an unsupported claim that male and female incident counts were similar. This portfolio edition contains one canonical notebook, one raw source file, a generated clean dataset, and explicit limitations.

Key findings

• Male incident records substantially outnumber female records (5,697 vs 814); the counts are not similar.
• Among records with known sex and outcome, fatality is 24.06% for male records and 16.78% for female records.
• Country and activity counts are strongly concentrated, but they are not risk rates because exposure denominators are unavailable.
• Long-run increases in recorded incidents mix real-world change with population, participation, communications, and record-coverage effects.
• Species information is frequently missing or unstructured, so species comparisons remain descriptive.

Repository structure

data/raw/global_shark_attacks.xls          Source register
data/processed/shark_incidents_clean.csv   Analysis-ready selected fields
notebooks/shark_incident_analysis.ipynb    Cleaning, validation, EDA and limits
outputs/                                   Exported summaries and charts
requirements.txt                           Reproducible environment

Reproduce the analysis

git clone https://github.com/vildanpir/Week-2-Project-Shark-Attacks-.git
cd Week-2-Project-Shark-Attacks-
python -m pip install -r requirements.txt
jupyter notebook notebooks/shark_incident_analysis.ipynb

Run the notebook from the repository root.

Interpretation boundary

This register records incidents, not time at risk. It cannot estimate the probability of an incident for a country, activity, sex, or species. The notebook deliberately uses terms such as “recorded incidents” and avoids causal or personal-safety claims.

Team

Vildan Pirpiroglu — structure, time, type, and location fields · Moci — victim and activity fields · Keagan — outcome and species fields