← All projectsSin Files / completed
Sin Files
Two cities. Two codebooks.
Shared fields. Separate analyses.
Houston960,326 · 2021–23
Los Angeles1,004,991 · 2020–
01Two codebooks, one language.
Different codebooks, mapped to shared fields.
LAPD
Houston · NIBRS
1845 → 18
- HOUR
- VICTIM_AGE
- VICTIM_SEX
- VICTIM_RACE
- VICTIM_ETHNICITY
- ucr_cat_id
02960,326 incidents. Four clusters.
Choose k on a 10,000-row sample. Fit four clusters to all Houston incidents. Dots show counts, not locations.
My part
I built the K-means and Random Forest branch and mapped LAPD fields to the project’s shared schema.
Team of two, with Jesus Mora-Manzo (spatial classification, association rules).
Shows
- Machine learning
- Feature scalingK-means clusteringSilhouette / Davies–BouldinRandom forests
- Data
- Schema harmonizationColumnar dataframesMissing-value imputationSampling for scale
Built with
PythonPolarspandasscikit-learnJupyter
Details
Method K-means · Random Forest · schema harmonization
Data Houston NIBRS, 2021–2023 (ICPSR) · LAPD Crime Data from 2020 to Present