Traffic accident is an event in the unanticipated and unintended involve vehicles with or without other road users, resulting in losses and/or loss of property. According Polrestabes Semarang number of traffic accidents decreased in 2014 compared to 2013, but the figure is still considered high. Therefore we need an analysis of traffic accident cases, in this case using a log linear models. Log linear models used to analyze the relationship between the response variables that are categories that make up the contingency table and determine which variables are likely to cause depedensi. In this study, the variable used is the severity of the victim, the type of accident, the role of the victim, the victim vehicle type, time of the accident and the age of the victim. The results indicate that the variables that affect the model is the severity of the victim, the type of accident, the role of the victim, the type of vehicle the victim, time of the accident, the age of the victim, the role of the victim * type of vehicle the victim, the type of accident * the role of the victim, the type of vehicle the victim * age of the victim, the type of accident * type of vehicle the victim, the severity of the victim * type of accident, type of accident * age of the victim. So that raises the most variable attachment is a type of accident.