2014 年 19 巻 1 号 p. 31-43
The objective of this study was to identify patterns of modus operandi that would be effective for predicting offender characteristics. For this purpose, logistic regression models were developed based on offender characteristics (17 variables) and modus operandi (50 variables) related to burglaries of residences, by using a stepwise method. Useful models were selected by using the area under the ROC curve, and by improvements in prediction accuracy from the prior probability, based on the frequency of the category of each offender characteristic. Models containing the following four offender characteristics were selected as useful: age, presence of a co-offender, criminal history of house burglaries with no one home, and criminal history of house burglaries late at night. Burglary experiences and age emerged as common potential factors of modus operandi and offender characteristics. It is suggested that selected modus operandi could be utilized to predict offender characteristics.