Predicting criminal offence in adolescents who exhibit antisocial behaviour: a machine learning study using data from a large randomised controlled trial of multisystemic therapy
Bibliographic record
Abstract
INTRODUCTION: Accurate prediction of short-term offending in young people exhibiting antisocial behaviour could support targeted interventions. Here we develop a set of machine learning (ML) models that predict offending status with good accuracy; furthermore, we show interpretable ML analyses can complement models to inform clinical decision-making. METHODS: This study included 679 individuals aged 11-17 years who displayed moderate-to-severe antisocial behaviour, from a controlled trial of Multisystemic-therapy in England. The outcome was any criminal offence in the 18 months after study baseline. Four types of ML algorithms were trained: logistic regression, elastic net regression, random forest, and gradient boosting machine (GBM). Prediction models were developed (1) using predictors readily available to clinicians (e.g. sociodemographics, previous convictions), and (2) with additional information (e.g. parenting). Model agnostic feature importance values were calculated and the most important predictors identified. Nested cross-validation with 100 iterations of random data splits and 10-fold cross-validation within each iteration was employed, and the average predictive performance was reported. RESULTS: Among the ML models using readily available predictors, the GBM is the strongest model (AUC 0.85, 95% CI 0.85-0.86); the other models have average AUCs of 0.82. This performance was better than using only the total number of previous offences as the predictor (0.67, 0.66-0.68), and the model simply assuming past offending status as the prediction (0.81, 0.80-0.81). Additional predictors slightly increased the performance of logistic regression and random forest models but decreased the performance of elastic net regression and gradient boosting machine-based models. CONCLUSION: The potential utility of ML approaches for accurately predicting criminal offences in high-risk youth is demonstrated. Interpretable ML-based predictive models could be utilised in youth services or research to help develop and deliver effective interventions.
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 | 0.030 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".