Predicting Arrest Release Outcomes: A Comparative Analysis of Machine Learning Models
Bibliographic record
Abstract
This comparative study evaluates machine learning models for predicting arrest release outcomes using 5,226 marijuana possession cases from the Toronto Police Service (1997-2002). The dataset exhibited significant class imbalance, with only 17.1% detention outcomes versus 82.9% releases. After preprocessing to handle missing values and convert categorical variables, we implemented two modeling approaches: a 500-tree Random Forest classifier with feature importance measurement and a binomial Logistic Regression model. Both algorithms demonstrated strong predictive capability for release cases, achieving comparable overall accuracy (83.2-83.4%) and excellent sensitivity (>98%), though they struggled with the critical minority class as evidenced by poor specificity (<7%). The models showed similar discriminative power, with Logistic Regression achieving a marginally higher AUC-ROC (0.733 vs 0.726). Feature importance analysis identified employment status and prior police background checks as the strongest predictors, while demographic factors, including race, also contributed significantly to predictions. These results highlight both the technical challenges of imbalanced classification in justice system data and the ethical considerations surrounding potential algorithmic bias, particularly given the high false positive rate for detention predictions that could exacerbate existing disparities. The study underscores the need for careful model evaluation and responsible implementation when applying predictive analytics to sensitive criminal justice decisions, balancing statistical performance with considerations of fairness and social impact.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.024 | 0.034 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.005 | 0.002 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".