conflict-data

Conflict hotspot detection applies DBSCAN clustering with Haversine distance to ACLED event coordinates to group political violence incidents into spatial clusters. You aggregate the raw event streams first, then use scikit-learn to identify geographic hotspots in kilometers.

Conflict onset logit modeling requires structuring ACLED or UCDP GED data into a country-year panel with lagged predictors and country fixed effects via dummies. You aggregate raw political violence events into panel summaries to fit the quantitative model.

UCDP GED fatality uncertainty handling structures low, best, and high battle-death estimate columns from the CSV data. You use pandas and geopandas to wrangle these specific uncertainty fields and summarize fatality trends over time.

Downloading ACLED conflict event data requires environment-based API credentials for proper authentication. You must configure these credentials in your environment before pulling raw political violence event streams via the API.

Aggregating UCDP GED conflict events into monthly time series patterns uses pandas to structure raw event streams into temporal summaries. You transform the discrete political violence incident data into consistent monthly counts for quantitative analysis.

Structuring raw political violence event streams for quantitative research involves using pandas and geopandas to wrangle ACLED or UCDP GED data into country-year and spatially clustered datasets. You obtain structured panels ready for conflict modeling and hotspot detection.