Data Mining Techniques for Recidivism Prediction: A Survey Paper

Recidivism has been the main concern of the law enforcements, namely the police officers, prosecutors, judges, correctional officials and parole boards in addressing crime rate. Recidivism describes the tendency of a convicted criminal to commit crime recurrently. A number of studies have been proposed for investigating and identifying factors of recidivism to overcome crime. Over the past few decades, we are called to answer the question, which technological solutions can best create the accurate predictive recidivism models. Given the complexity of the recidivism data, the emergence of data mining techniques proves that it is one of the sought after tools to analyze data patterns. The aim of this paper is to review several researchers on recidivism, which apply data mining techniques to work with complex recidivism data. We review five data mining techniques used for prediction of recidivism. Clustering, Association Rules, Decision Tree, Support Vector Machine and Neural Networks are among the selected techniques. However, this review does not cover all the data mining techniques, but to only emphasize based on popularity and most exploited method over the last two decades. The findings of this review shows that data mining techniques have been applied in recidivism research apart from the standard parametric (logistic regression) and actuarial risk assessment models. Nevertheless, there is a drawback in using data mining tools that is the absence of the explanatory variables in recidivism data set.