A Resarch on the Factors Affecting the Outcomes of Child Abuse Cases Using Machine Learning Methods

Modern information technology makes it possible to collect and store scientific and social research data. Some statistical methods can provide quite reliable results when the necessary assumptions are met in uncovering existing or hidden relationships between data. However, since data collected from real life often do not meet these assumptions, data mining methods that require fewer assumptions and can be applied to flexible and complex data sets have been developed for prediction. The use of machine learning methods, which include data mining techniques, to process data and produce meaningful information has become widespread in recent years. In this study, techniques such as the CHAID algorithm, an application of decision trees, and support vector machines, were compared with the logistic regression analysis method. The study’s sample consists of data from 61 child abuse cases in which the UCIM Saadet Öğretmen Association Struggling Child Abuse requested participation. The dependent variable of the study is whether the defendant received a sentence at the end of the trial, while the independent variables are five variables identified by leveraging expert (lawyer) opinions. As a result, it was found that the CHAID algorithm and support vector machines provided more accurate classification.