Author:
Kovalchuk Olha,Banakh Serhiy,Chudyk Nataliia,Drakokhrust Tetiana
Abstract
The modern information society requires effective digital justice based on innovative technologies. This research aims to create machine-learning algorithms to evaluate the likelihood of prisoners reoffending, utilising their socio-demographic attributes and past criminal history. In this paper, the experimental method, modelling method, forecasting, field research, statistical analysis, case study, meta-analysis, comparative analysis, and machine learning techniques have been used. It was investigated that the main factors influencing the risk level (low, moderate, high) of recidivism are dynamic characteristics associated with previous criminal activities (court decisions for specific individuals provided for suspended sentences and early releases, rather than serving sentences in correctional institutions). The age at which a person was first involved in the criminal environment (first convicted to a suspended sentence or imprisonment for a certain period while serving in correctional institutions) also significantly affects the propensity for criminal relapse. Individual characteristics of convicts (age at the time of the study, gender, marital status, education level, place of residence, type of employment, motivation for release) are not correlated with a tendency to commit repeated crimes. The age at which a person was first sentenced to actual imprisonment or given their first suspended sentence, the age at which a person was first sentenced to the actual degree of punishment, the number of early dismissals, and the young age at which a person was first involved in the criminal environment (received their first suspended conviction or real conviction) are significant factors increasing the risk of committing a recidivist crime in the future. The proposed model can be applied to predict the level of propensity for recidivism crimes for new cases. The obtained results can provide reliable information support for court decisions and become part of a comprehensive court information system
Publisher
Scientific Journals Publishing House
Reference28 articles.
1. [1] 15 must-know machine learning algorithms. A comprehensive guide for machine learning. (2021). Retrieved from https://towardsdatascience.com/15-must-know-machine-learning-algorithms-44faf6bc758e.
2. [2] Alcántara Francia, O.A., Nunez-del-Prado, M., & Alatrista-Salas, H. (2022). Survey of text mining techniques applied to judicial decisions prediction. Applied Sciences, 12(20), article number 10200. doi: 10.3390/app122010200.
3. [3] Alghazzawi, D., Bamasag, O., Albeshri, A., Sana, I., Ullah, H., & Asghar, M.Z. (2022). Efficient prediction of court judgments using an LSTM+CNN neural network model with an optimal feature set. Mathematics, 10(5), article number 683. doi: 10.3390/math10050683.
4. [4] Anazodo, K.S., Ricciardelli, R., & Chan, Ch. (2019). Employment after incarceration: Managing a socially stigmatized identity. Equality, Diversity and Inclusion, 38(5), 564-582. doi: 10.1108/EDI-09-2018-0175.
5. [5] Berezka, K.M., Kovalchuk, O.Ya., Banakh, S.V., Zlyvko, S.V., & Hrechaniuk, R. (2022). A binary logistic regression model for support decision making in criminal justice. Folia Oeconomica Stetinensia, 22(1), 1-17. doi: 10.2478/foli-2022-0001.