MSF Forecast — Weekly U.S. Mass-Shooting Incidence on MSF Forecast — Mass-Shooting Forecasting Project

Data & Reproducibility

Sat, 11 Jul 2026 00:00:00 +0000

Reproduction (the actual code, end-to-end)

git clone https://git.catalystgroup.tech/herman/msf-forecast-pipeline
cd msf-forecast-pipeline
pip install -r requirements.txt # numpy, pandas, matplotlib, scipy, statsmodels, pytest
make all # ≈ 2 minutes on a single CPU

End-to-end runtime: ~2 minutes on a single CPU. Six pytest cases cover the loaders, feature builder, and metric helpers. Output:

results/features.csv — 522-week × 40-feature panel
results/predictions_test.csv — 104-week × 20-col TEST-set forecasts + actuals
results/metrics_{count,any,high}.csv — TEST-set metrics per task
results/cv_metrics.csv — rolling-origin development-CV scores
results/model_coefficients.csv — full NB GLM coefficients
results/model_metadata.json — study-weeks, RF-skip reason, training window
results/calibration_any.png — reliability diagram for the any-incident target
results/figures/{observed_vs_predicted,feature_importance,residuals,roc_any}.png

Inputs (all public, all committed under `src/data/raw/` in the pipeline repo)

File	Rows	Source	Span
`gva_mass_shootings.csv`	82	Codeholics/US-Mass-Shootings (Mother Jones enrichment)	2014–2023
`gtrends_mass_shooting.csv`	120	proxy: monthly incident count from the same source	2014-01 → 2023-12
`nics_monthly.csv`	299	BuzzFeed News / FBI NICS	1998-11 → 2023-09
`us_population.csv`	11	World Bank `SP.POP.TOTL` (USA)	2014–2024

To regenerate from the original public sources:

Figures

Sat, 11 Jul 2026 00:00:00 +0000

All five figures are downloaded directly from the pipeline repo at results/figures/ (commit cf180acf). Click to open full-size PNG.

Observed vs Predicted (TEST window)

The full-feature negative-binomial GLM point predictions against actual weekly counts for the held-out 2022-W01 → 2023-W52 window.

![](](/results/figures/observed_vs_predicted.png)

Feature Importance

Top-20 features by absolute coefficient from the full-feature NB GLM, ordered for stability across re-fits.

![](](/results/figures/feature_importance.png)

Residuals (TEST)

Time series of (y - y_hat) on the test window, with shaded ±1σ bands from the rolling-origin backtest residuals.

Methodology

Sat, 11 Jul 2026 00:00:00 +0000

Targets

Three weekly targets are forecast independently:

target_count — Poisson / negative-binomial regression of weekly incident counts. Range 0–3 in TEST.
target_any (binary, ≥1 incident in week) — logistic regression. Base rate in the TEST window: 22% (23 of 104 weeks).
target_high (binary, ≥2 incidents in week) — logistic regression. Base rate in the TEST window: 1% (1 of 104 weeks).

All targets share the same weekly panel and the same feature set.

Paper

Sat, 11 Jul 2026 00:00:00 +0000

Abstract

This study tests whether a small collection of publicly observable signals can forecast weekly United States public mass-shooting incidence. The outcome is the number of Mother Jones / Codeholics-defined incidents in each Monday–Sunday ISO week: a public, indiscriminate shooting with at least three fatalities. We build a 522-week panel from 2014-W01 through 2023-W52 and reserve the final 104 weeks, 2022-W01 through 2023-W52, as a fixed test set. Predictors include trailing incident counts, seasonality, holiday and school-session indicators, lagged FBI NICS counts, annual population, and a completed-month incident-frequency proxy for public attention. We compare a week-of-year seasonal mean, a lag-only negative-binomial (NB) model, a regularized full NB model, and regularized logistic models for "any incident" and "at least two incidents."

Results

Sat, 11 Jul 2026 00:00:00 +0000

TEST window: 2022-W01 → 2023-W52 (104 weeks, held out from all model fitting). TEST window observed: 23 weeks with ≥1 incident (prevalence 22%), 1 week with ≥2 incidents (prevalence 1%).

Count task — best on tail-quantile loss

Model	RMSE	MAE	MAPE	pinball 0.5	pinball 0.8	pinball 0.9
naive_seasonal	0.4653	0.3096	30.00	0.1154	0.1865	0.1087
nb_lag	0.4524	0.3191	31.02	0.1154	0.1942	0.0827
nb_full	0.4502	0.3438	33.50	0.1154	0.1673	0.0817

The full-feature Negative-Binomial model (nb_full) dominates the upper quantiles (pinball@0.8 and pinball@0.9) where mass-shooting forecasting actually matters. RMSE / MAE / MAPE are near-tied across all three models — the target class is sparser than the autoregressive signal can resolve.

MSF Forecast — Weekly U.S. Mass-Shooting Incidence on MSF Forecast — Mass-Shooting Forecasting Project

Data & Reproducibility

Reproduction (the actual code, end-to-end)

Inputs (all public, all committed under src/data/raw/ in the pipeline repo)

Figures

Observed vs Predicted (TEST window)

Feature Importance

Residuals (TEST)

Methodology

Targets

Paper

Abstract

Results

Count task — best on tail-quantile loss

Inputs (all public, all committed under `src/data/raw/` in the pipeline repo)