Improved Crime Prediction Using Hybrid Neural Architecture Search Together with Hyperparameter Tuning
Bibliographic record
Abstract
Different parts of the world have recorded an escalating number of criminal incidents, burdened the judicial system, and adversely impacted national security and economic development. Accurate prediction of crimes is crucial for law enforcement agencies to prevent proactively criminal activity and allocate resources effectively. Existing methods often address architecture design and hyperparameter tuning as separate processes. We present a combination of neural architecture search and hyperparameter tuning for improved crime prediction. The study method achieved automation of architecture discovery and fine-tuning of hyperparameters by utilizing Neural Architecture Search (NAS) to explore a wide range of neural network architectures for crime prediction and optimizing the hyperparameters of the discovered architecture for peak performance in binary crime prediction, respectively. The study used three datasets: criminal cases dataset (self-collected dataset), Vancouver crime data, and Austin crime data. The criminal cases dataset is extracted from a confidential database from certain countries, focusing on sensitive parts of those countries. The Vancouver crime and Austin crime datasets were sourced from the Kaggle website. The study considered the robust rank aggregation (RRA) feature selection method to rank and select the best features to predict crime behavior in some countries. The chosen features using robust rank aggregation included current position, age range, month, prisoner condition, and identified/unidentified (ide/unide). The hyperparameter tuning model of Architecture Search (NAS +) produced superior results across all datasets with an accuracy of 89.29% (AUC-ROC = 94.82% and recall = 64.54%) in the criminal cases dataset, 60.37% (AUC-ROC = 50.00% and recall = 100.00%) in the Vancouver dataset, and 86.68% (AUC_ROC = 65.40% and recall = 100.00%) in the findings which demonstrated that the proposed approach consistently outperforms conventional methods, making it an effective solution for the prediction of real-world crimes.
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 | 0.002 | 0.006 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".