Predicting Recidivism among Inmates Population using Artificial Intelligent (AI) Techniques: A Case study of Kenya Prisons Department

Gikaru, W. Judy (2014) Predicting Recidivism among Inmates Population using Artificial Intelligent (AI) Techniques: A Case study of Kenya Prisons Department. Masters thesis, University of Nairobi.

PDF (Predicting Recidivism among Inmates Population using Artificial Intelligent (AI) Techniques: A Case study of Kenya Prisons Departmen) Gikaru_Predicting recidivism among inmates population using Artificial Intelligent (AI) techniquest(1).pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Currently in the Kenya prisons department there is no defined way of checking the rate of recidivism among the prison inmates population. The officers rely only on manual tallying of prisoners during admission which is not efficient. With the increase use of computerized systems in the department there is need to implement those that can help in rehabilitation and reformation. In this research Artificial intelligent techniques that is decision tree, neural networks and bayesnets are used to check on the rate of recidivism in the inmate’s population. This is illustrated by the development of the Recidivism Prediction System (RPS) prototype, using the WEKA tool and the python GUI application, which play a major role in risk assessment of the inmates by checking their rate of recidivism. Currently congestion in the prisons institutions is a major challenge to the management, since the resources provided doesn’t match up the need on the ground. Using the RPS prototype the department management can be able to visualize various patterns on recidivism from predicted result and most importantly show the prisoners likely rate of recidivism. Assisting the users in the decision making process, as rehabilitation and reformation is not just about incarnation but also include Community Service Order and parole. The RPS prototype is important to the users as it can be used to predict recidivism rate and plan on various programs on rehabilitation and reformation to introduce or not. As from the prototype results the prediction outcome vary from one instance to another, where those with value above TWICE are of higher recidivism risk compared to those with ONCE and below. The prediction results is also compared with other attributes and displayed for better understanding.

Item Type:	Thesis (Masters)
Subjects:	Q Science > Q Science (General) Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:	Africana
Depositing User:	Mr Jude Abhulimen
Date Deposited:	22 Jul 2016 08:06
Last Modified:	22 Jul 2016 08:14
URI:	http://thesisbank.jhia.ac.ke/id/eprint/795

Actions (login required)

View Item