MU Digital Repository
Logo

Supervised Machine Learning Approach to Predict Crime Recidivism

Dhankar, Abhishek and Sandhu, Arshdeep (2017) Supervised Machine Learning Approach to Predict Crime Recidivism. In: Supervised Machine Learning Approach to Predict Crime Recidivism, 01/07/2017, Manipal Institute of Technology, Manipal.

[img] PDF
801.pdf - Published Version
Restricted to Registered users only

Download (584kB) | Request a copy

Abstract

Recidivism Prediction System can help in predicting the chances of a convict committing crime again, which would enable the judges to take informed decision on the severity of punishment or imposing alternative punishment like compulsory social service without sending a convict to jail, depending upon the likelihood whether a convict will commit crime again or not. This will also help in taking decision on early release of prisoners who are not likely to re-offend and longer incarceration for those who are likely to re-offend, thereby reducing the overcrowding in prisons and government expenditure, better rehabilitation of convicts without compromising on public safety. In this paper, we have proposed a supervised machine learning approach to predict the likelihood of recidivism. Initially, some parameters related to individual crime records were chosen to build training data set and these parameters were trained using Artificial Neural Network with back propagation algorithm. This back propagation learning process iteratively changes or optimizes the parameters during training phase in the Artificial Neural Network. Finally, the predictions were made for the test data on the basis of the parameters calculated on the basis of training data. In the present study, several sets of training and test data are taken from the Govt. website of Iowa State for implementation and improved results have been obtained.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: Recidivism, Supervised machine learning, Artificial Neural Network, back propagation learning
Subjects: Engineering > MIT Manipal > Computer Science and Engineering
Depositing User: MIT Library
Date Deposited: 09 Aug 2017 03:45
Last Modified: 09 Aug 2017 03:45
URI: http://eprints.manipal.edu/id/eprint/149482

Actions (login required)

View Item View Item