GPS RESEARCH LIBRARY: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) (NCJ 305039, 2021) ============================================================ Georgia Prisoners' Speak — gps.press Generated: 2026-09-14 13:24:15 EDT Research Date: 2026-09-06 Topic: Recidivism & Reentry JSON: https://gps.press/research-data/predicting-criminal-recidivism-using-specialized-feature-engineering-and-xgboost-nij-recidivism-forecasting-challenge-georgia-parolees-released-2013-ncj-305039-2021/?format=json SUMMARY ---------------------------------------- This NIJ-funded winning paper applies XGBoost machine learning to predict recidivism among approximately 26,000 individuals released from Georgia prisons on discretionary parole in 2013. The authors achieved best-in-class Brier scores for Year 2 recidivism prediction, with the 'Early Recidivism' feature (Recidivism_Arrest_PrevYear) identified as a key driver of model performance. The paper documents feature importance rankings, model parameters, and performance metrics, providing methodological insights for GPS research on Georgia parolee recidivism forecasting. STATISTICS (15) ---------------------------------------- - [reported] Dataset size: ~26,000 Georgia parolees released 2013 The aggregated dataset provided by the NIJ contains approximately 26,000 individuals released from Georgia prisons on discretionary parole for post-incarceration supervision between January 1st, 2013 and December 31st, 2013. Value: 26000.0 individuals Date: 2013-01-01 Tags: parole,demographics,facilities Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Year 1 model Brier score 0.1837 Year 1 model achieved a Brier score of 0.1837 and F1 score of 0.3009. Value: 0.1837 Brier score Date: 2021-01-01 Tags: recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Year 2 model Brier score 0.1172 Year 2 model achieved a Brier score of 0.1172 and F1 score of 0.2512. Value: 0.1172 Brier score Date: 2021-01-01 Tags: recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Year 3 model Brier score 0.0720 Year 3 model achieved a Brier score of 0.0720 and F1 score of 0.0492. Value: 0.072 Brier score Date: 2021-01-01 Tags: recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Neural Network 10-fold validation Brier score mean 0.1525 Neural Network with hyperparameter tuning achieved a mean Brier score of 0.1525 (SD 0.00336) and mean F1 score of 0.7850 (SD 0.0090) in 10-fold validation. Value: 0.1525 Brier score Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Random Forest 10-fold validation Brier score mean 0.1087 Random Forest with hyperparameter tuning achieved a mean Brier score of 0.1087 (SD 0.000126) and mean F1 score of 0.8390 (SD 0.00152) in 10-fold validation. Value: 0.1087 Brier score Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost 10-fold validation Brier score mean 0.0945 XGBoost with hyperparameter tuning achieved a mean Brier score of 0.0945 (SD 0.000335) and mean F1 score of 0.8559 (SD 0.00128) in 10-fold validation. Value: 0.0945 Brier score Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Oracle team 1st place Year 2 female parolees Brier score 0.1233 In Year 2 female parolees competition, Oracle placed 1st with a Brier score of 0.1233. Value: 0.1233 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] MCHawks team 2nd place Year 2 female parolees Brier score 0.1242 In Year 2 female parolees competition, MCHawks placed 2nd with a Brier score of 0.1242. Value: 0.1242 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] VT-ISE team 3rd place Year 2 female parolees Brier score 0.1260 In Year 2 female parolees competition, VT-ISE placed 3rd with a Brier score of 0.1260. Value: 0.126 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] DEAP team 4th place Year 2 female parolees Brier score 0.1263 In Year 2 female parolees competition, DEAP placed 4th with a Brier score of 0.1263. Value: 0.1263 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] MCHawks team 1st place Year 2 male & female parolees Brier score 0.1405 In Year 2 male & female parolees competition, MCHawks placed 1st with a Brier score of 0.1405. Value: 0.1405 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Oracle team 2nd place Year 2 male & female parolees Brier score 0.1451 In Year 2 male & female parolees competition, Oracle placed 2nd with a Brier score of 0.1451. Value: 0.1451 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] VT-ISE team 3rd place Year 2 male & female parolees Brier score 0.1472 In Year 2 male & female parolees competition, VT-ISE placed 3rd with a Brier score of 0.1472. Value: 0.1472 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] DEAP team 4th place Year 2 male & female parolees Brier score 0.1481 In Year 2 male & female parolees competition, DEAP placed 4th with a Brier score of 0.1481. Value: 0.1481 Brier score Date: 2021-01-01 Tags: recidivism,methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) METHODOLOGY NOTES (17) ---------------------------------------- - [reported] NIJ train/test split proportion 70/30 NIJ split the dataset into a training and test set with a 70/30 proportion. Date: 2021-01-01 Tags: methodology,operations Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Data providers: GDCS and Georgia Bureau of Investigation Both the GDCS and the Georgia Bureau of Investigation provided data. The GDCS data included demographics, prison and parole case information, prior community supervision history, and supervision activities. The Georgia Bureau of Investigation provided data on prior criminal history measures such as arrest and conviction episodes prior to prison entry. Date: 2013-01-01 Tags: methodology,operations,parole Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Recidivism measure definition: new felony or misdemeanor arrest within 3 years GCIC data also provides the recidivism measure, defined as a new felony or misdemeanor arrest episode within three years of parole supervision start date. This recidivism measure includes three dichotomous variables measuring if an individual recidivated in the three-year follow-up period (yes/no) as well as recidivated by time period (year 1, year 2, or year 3). Tags: recidivism,methodology,parole Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Early Recidivism feature added for Year 2 and Year 3 models For Year 2 and Year 3, a feature called Early Recidivism was added. This feature delineated whether the individual already returned to prison. This could only be added to Year 2 training and Year 3 training. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Low feature importance variables retained in model There were multiple variables that had relatively low feature importance: Gender, Prior_Conviction_Episodes_PPViolationCharge, Prior_Conviction_Episodes_DomesticViolenceCharges, to name a few. However, because computational speed wasn't a priority for this project, all variables, regardless of statistical significance, were kept in model training. Date: 2021-01-01 Tags: methodology,demographics Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Software: Python 3.8 and scikit-learn Preprocessing and model construction were performed on Python 3.8. Preprocessing functions and ML models were imported from the Python library scikit-learn. Date: 2021-01-01 Tags: methodology,operations Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Missing values imputed using SimpleImputer Missing values were imputed using the SimpleImputer library. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Team entered Small Team category of NIJ challenge Our team entered into the Small Team category of the challenge and aimed to utilize state of the art machine learning techniques to assist in this field. Date: 2021-01-01 Tags: methodology,operations Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: colsample_bytree 0.75 Optimized XGBoost parameter colsample_bytree set to 0.75. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: learning_rate 0.1456 Optimized XGBoost parameter learning_rate set to 0.1456. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: max_depth 6 Optimized XGBoost parameter max_depth set to 6. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: min_child_weight 2 Optimized XGBoost parameter min_child_weight set to 2. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: n_estimators 925 Optimized XGBoost parameter n_estimators set to 925. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] XGBoost parameters: subsample 0.7 Optimized XGBoost parameter subsample set to 0.7. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Grid search hyperparameter tuning used for XGBoost Parameters of the XGBoost were optimized using grid search hyperparameter tuning. Parameters that were optimized for are number of estimators, learning rate, and max depth of XGBoost trees. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] 10-fold validation performed for each model For each model trained, 10-fold validation was performed to measure average performance. Metrics captured were brier score and F1 score. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Categorical variables one-hot-encoded, ordinal integer-encoded Variables were split into categorical, ordinal, and numerical. Categorical variables were one-hot-encoded and ordinal variables were integer-encoded. Numerical variables were scaled to be between 0 and 1. Date: 2021-01-01 Tags: methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) FINDINGS (12) ---------------------------------------- - [reported] XGBoost selected as best performing model In Round 1 of the competition, we analyzed different model performances using 10 fold cross validation and found that on average XGBoost performed the best. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Year 2 model performed very well on hidden dataset In terms of our performance in the challenge, we found that our Year 2 model performed very well on the hidden dataset when compared with other submissions. Date: 2021-01-01 Tags: recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Random Forest performed well but more variable than XGBoost The Random Forest performed well but had a much more variable performance than the XGBoost. Other regression and classification methods were assessed but performances varied widely, none achieving the success of the XGBoost. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Class imbalance affected all models All models suffered from heavy class imbalance due to the relative proportion of the non-recidiviated class to the recidivated class. Synthetic augmentation methods like SMOTE were used to mitigate this class imbalance, but did not provide significant benefits. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Optimal probability threshold 0.5 for XGBoost Different probability thresholds were experimented to optimize XGBoost predictions. With the training and testing data split that was used, optimal brier scores were achieved by using a 0.5 threshold. Increasing or decreasing the threshold leads to worse metrics. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Early Recidivism feature significantly improved model performance More specifically the addition of our Early Recidivism feature (Recidivism_Arrest_PrevYear) played a significant role in helping our model performance improve for the later rounds when compared with other submissions. Date: 2021-01-01 Tags: recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] F1 score recommended for future imbalanced data studies One metric that could be used in future studies is the F1 score. The F1 score is the harmonic mean of the precision and recall, making it an effective score for delineating the performance of models which were trained on highly imbalanced data. Date: 2021-01-01 Tags: methodology,recidivism Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Top 5 features for Year 1 recidivism prediction Top five most important features for prediction of recidivism in Year 1: Age_at_Release, Prior_Arrest_Episodes_Felony, Gang_Affiliated, Prison_Years, Prior_Arrest_Episodes_Property. Date: 2021-01-01 Tags: recidivism,gangs,demographics,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Top 5 features for Year 2 recidivism prediction Top five most important features for prediction of recidivism in Year 2: Recidivism_Arrest_PrevYear, Percent_Days_Employed, Jobs_Per_Year, Age_at_Release, Avg_Days_per_DrugTest. Date: 2021-01-01 Tags: recidivism,employment,drugs,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Top 5 features for Year 3 recidivism prediction Top five most important features for prediction of recidivism in Year 3: Recidivism_Arrest_PrevYear, Percent_Days_Employed, Jobs_Per_Year, Age_at_Release, Avg_Days_per_DrugTest. Date: 2021-01-01 Tags: recidivism,employment,drugs,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] COMPAS system identified as early criminal justice AI tool One of the earliest tools developed that utilized these new technologies was the Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) system by Northpoint. Tags: methodology,policy,operations Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) - [reported] Models can supplement existing recidivism technologies While the models built suffer from class imbalance, data points with a high probability of being positive for recidivism are likely going to exhibit recidivism in reality. Hence, the models can be used as a supplement to existing technologies in the recidivism space. Date: 2021-01-01 Tags: recidivism,methodology,policy Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) POLICYS (1) ---------------------------------------- - [reported] NIJ Recidivism Forecasting Challenge goals With this Challenge, NIJ aims to: 1) encourage 'non-criminal justice' forecasting researchers to compete against more 'traditional' criminal justice forecasting researchers, building upon the current knowledge base while infusing innovative, new perspectives; and 2) compare available forecasting methods in an effort to improve person-based and place-based recidivism forecasting. Date: 2021-01-01 Tags: policy,recidivism,methodology Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) LEGAL FACTS (1) ---------------------------------------- - [reported] Recidivism definition per US Department of Justice Recidivism is defined by the US Department of Justice as 'a person's relapse into criminal behavior, often after the person receives sanctions or undergoes intervention for a previous crime'. It is measured by the criminal acts that result in rearrest, reconviction or return to prison in the three year period following a prisoner's release. Tags: recidivism,legal,policy Sources: Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013) DATASETS (6) ---------------------------------------- # XGBoost Model Performance Metrics by Year Brier score and F1 score for XGBoost models predicting recidivism in Years 1, 2, and 3 for Georgia parolees released in 2013. Model Brier Score F1 Score ------------------------------- Year 1 0.1837 0.3009 Year 2 0.1172 0.2512 Year 3 0.072 0.0492 # Optimized XGBoost Hyperparameters Final hyperparameter values used for XGBoost models in the NIJ Recidivism Forecasting Challenge. Parameter Value -------------------------- Colsample_bytree 0.75 learning_rate 0.1456 max_depth 6 min_child_weight 2 n_estimators 925 subsample 0.7 # Ten-Fold Validation Scores of Different Models Mean and standard deviation of Brier score and F1 score across 10-fold validation for Neural Network, Random Forest, and XGBoost models with hyperparameter tuning. Model Brier Score Mean Brier Score SD F1 Score Mean F1 Score SD ------------------------------------------------------------------------------ Neural Network 0.1525 0.00336 0.785 0.009 Random Forest 0.1087 0.000126 0.839 0.00152 XGBoost 0.0945 0.000335 0.8559 0.00128 # Year 2 Female Parolees Competition Rankings Brier scores and rankings for teams in the Year 2 female parolees recidivism prediction competition. Place Team Name Brier Score ------------------------------- 1 Oracle 0.1233 2 MCHawks 0.1242 3 VT-ISE 0.126 4 DEAP 0.1263 # Year 2 Male & Female Parolees Competition Rankings Brier scores and rankings for teams in the Year 2 male and female parolees recidivism prediction competition. Place Team Name Brier Score ------------------------------- 1 MCHawks 0.1405 2 Oracle 0.1451 3 VT-ISE 0.1472 4 DEAP 0.1481 # Top Five Most Important Features for Recidivism Prediction by Year Ranked list of the top five most important features for predicting recidivism in Years 1, 2, and 3. Year 1 Year 2 Year 3 ---------------------------------------------------------------------------------------- Age_at_Release Recidivism_Arrest_PrevYear Recidivism_Arrest_PrevYear Prior_Arrest_Episodes_Felony Percent_Days_Employed Percent_Days_Employed Gang_Affiliated Jobs_Per_Year Jobs_Per_Year Prison_Years Age_at_Release Age_at_Release Prior_Arrest_Episodes_Property Avg_Days_per_DrugTest Avg_Days_per_DrugTest KEY ENTITIES (13) ---------------------------------------- - COMPAS [program]: One of the earliest criminal justice risk assessment tools utilizing data analytics and AI, developed by Northpoint. (aka: Correctional Offender Management Profiling for Alternative Sanctions) - DEAP [organization]: Competing team in the NIJ Recidivism Forecasting Challenge, placed 4th in both Year 2 female parolees and Year 2 male & female parolees. - Georgia Bureau of Investigation [organization]: Georgia state law enforcement agency that conducts some criminal investigations involving the prisons (aka: GBI) - Georgia Crime Information Center [organization]: State criminal justice information system connected to the national CJIS database, used for warrant entry and wanted person identification. (aka: GCIC) - Georgia Department of Corrections [organization]: State agency responsible for operating Georgia's prison system. Subject of federal DOJ investigation in 2022-2023 for constitutional violations including food-related deaths. (aka: GDC) - MCHawks [organization]: Competing team in the NIJ Recidivism Forecasting Challenge, placed 2nd in Year 2 female parolees and 1st in Year 2 male & female parolees. - National Institute of Justice [organization]: DOJ research agency that commissioned studies on deaths in custody to meet DCRA 2013 requirements. (aka: NIJ) - Northpoint [organization]: Developer of the COMPAS risk assessment system. - Oracle [organization]: Competing team in the NIJ Recidivism Forecasting Challenge, placed 1st in Year 2 female parolees and 2nd in Year 2 male & female parolees. - Prathic Sundararajan [person]: Co-author of the winning paper for the NIJ Recidivism Forecasting Challenge. - Suraj Rajendran [person]: Co-author of the winning paper for the NIJ Recidivism Forecasting Challenge. - US Department of Justice [organization]: Federal department that defines recidivism and funded the NIJ Recidivism Forecasting Challenge. (aka: Department of Justice, DOJ) - VT-ISE [organization]: Competing team in the NIJ Recidivism Forecasting Challenge, placed 3rd in both Year 2 female parolees and Year 2 male & female parolees. SOURCES (6) ---------------------------------------- - Class imbalanced Learning Menggunakan Algoritma SYNTHETIC MINORITY OVER-SAMPLING Technique – Nominal (smote-n) Pada dataset Tuberculosis anak, Jurnal Buana Informatika by Kurniawati, Y. E. (2019-01-01) [academic, primary] URL: https://doi.org/10.24002/jbi.v10i2.2441 - NIJ's Definition of Recidivism, National Institute of Justice by Department of Justice [official_report, primary] URL: https://nij.ojp.gov/topics/corrections/recidivism - Predicting Criminal Recidivism Using Specialized Feature Engineering and XGBoost (NIJ Recidivism Forecasting Challenge, Georgia parolees released 2013), NCJRS Virtual Library / Office of Justice Programs by Suraj Rajendran, Prathic Sundararajan (2022-07-01) [academic, primary] URL: https://www.ojp.gov/pdffiles1/nij/grants/305039.pdf - Recidivism Forecasting Challenge, National Institute of Justice by Department of Justice (2021-01-01) [official_report, primary] URL: https://nij.ojp.gov/funding/recidivism-forecasting-challenge - Setting the Record Straight: What the COMPAS Core Risk and Need Assessment Is and Is Not, Harvard Data Science Review by Jackson, E., & Mendoza, C. (2020-01-01) [academic, secondary] URL: https://doi.org/10.1162/99608f92.1b3dadaa - Xgboost: A Scalable Tree Boosting System, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining by Chen, T., & Guestrin, C. (2016-01-01) [academic, primary] URL: https://doi.org/10.1145/2939672.2939785