- Cleaned and standardized city, state, country, and time fields.
- Explored frequency of shapes, durations, and reporting trends.
- Chi-square test & Cramér’s V used to assess shape correlations by location.
Using GeoPandas, sightings were plotted on a global basemap to explore density by region.
Analyzed yearly, monthly, weekday, and hourly sighting patterns.
- Most sightings occur in summer, especially July.
- Reports spike between 8 PM and midnight.
- Sightings peaked around 1997–2003.
Applied text cleaning, sentiment analysis, and word cloud generation on over 80,000 sighting descriptions.
- Majority of reports are neutral or mildly positive.
- Language tends to be objective, focusing on observation.
- Word cloud of most common terms below:
- Python (Pandas, NumPy, Matplotlib, Seaborn, GeoPandas)
- NLP:
TextBlob,wordcloud,nltk - Geospatial:
geopandas,shapely,geodatasets - Jupyter Notebooks
- Topic modeling on sighting descriptions
- Time-series forecasting or anomaly detection
- Interactive dashboard (e.g., with Plotly Dash or Streamlit)
MIT
Data from NUFORC via Kaggle.