AltReform Policy Predictor: Criminal Justice Reform Initiatives, 1995-2025

578 historical criminal justice reform initiatives from 1995 to 2025, encoded as feature vectors with binary outcome labels. The training corpus and code support a four-model ensemble (Logistic Regression, Random Forest, XGBoost, SVM-RBF) for predicting reform outcomes. The dataset was authored by Joseph McLean-Arthur and is hosted on Harvard Dataverse.

Use Cases

• Train a machine learning ensemble to predict the success probability of policy initiatives based on 38 encoded features.
• Reproduce SHAP feature importance analysis to identify key factors influencing criminal justice reform outcomes.
• Benchmark the performance of different classification algorithms (Logistic Regression, Random Forest, XGBoost, SVM-RBF) on a structured policy dataset.
• Analyze historical trends in reform initiatives and their outcomes over a 30-year period.

Strengths

• Contains 578 historical reform initiatives, providing a substantive sample for analysis.
• Each initiative is encoded with 38 features, offering a detailed vector representation.
• Covers a 30-year time range (1995-2025) for longitudinal study.
• Includes reproducible code for model training, evaluation, and SHAP analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but the specific file formats and data size are unknown, which may limit suitability assessment.

Provenance

Source

Harvard Dataverse

Collection Method

Encoded from published meta-analyses and program evaluations.

Time Range

1995-2025

Freshness

Last updated 2026-06-09 15:24:46; freshness should be verified.