범죄자 프로파일링을 위한 성적살인 연구: 랜덤포레스트(Random Forest)를 활용하여

Current study is conducted to explore characteristic of sexual homicide in Korea, and to present model which shows best explanation about sexual homicide. Current study uses random forest analysis that is a one of method by machine learning technic to solve methodological problem of preceding studies, and logistic regression to present predicting direction of variables. Current study inputs variables related with offender, victim, and crime scene behaviors. Result in study, our model has moderate level of predicting indicator somewhat, this predictor will be greater by adding cases and information. Relationship with offender and victim(non relationship) is the best predicting variable, and the lowest is location of body discover. Result in regression, offender’s sex(men) and marriage status(single) predict sexual homicide positively, and education level predicts sexual homicide negatively. When offender did not have relationship with victim, overkill predicts sexual homicide negatively. Current study has some implications in terms of exploring sexual homicide model to help crime investigation and offender profiling. Also we expect that current study can expand study of sexual homicide in Korea. Current study has academical implication in terms of application for machine learning technic to analysis of sexual homicide.