In recent years, the problem of runaway youths, which is rapidly increasing and worsening in nature, has emerged a serious problem to the individual, home and the society as a whole. Against this back drop, this study is designed to shed light on the danger signals through an extensive analysis of the factors influencing the running away of youths and on this basis build a forecast model on runaway youths. A questionnaire was distributed among 906 subjects living in Seoul and Kyunggido, among 843 of whose answers have analysed through SAS. The differences between the runaway group and the non-runayay group were subject to chi-square and t-test. Also logistic regression analysis was conducted on the basis of purposeful selection method for constructing the forecast model. Further, canonical discriminant analysis including explanatory variables selected in the final model was undertaken to evaluate the discrimination of the model. The findings are as follows : the major predicting factors of runaway youth are examination scores, sexual experience, and antisocial personality. Home-related factors were found to be the absence of either or both of parents, running away of siblings, overprotective or highly goal-oriented rearing. Society-related factor was identified as the existence of a friend or friends of the opposit sex. The goodness-of-fit according to the preeditiorial variables, i.d. -2LL was 519.308 while Model chi-square turned out to be 439.558(df=9, p=0.0001), which left the sensitivity at 82.4% and the specificity at 82.6%.