The ubiquity of vending machines has led to numerous misdemeanors involving them, including much vandalism committed as pranks, or under the influence of alcohol, but also including organized professional crime, to which few effective countermeasures have been devised. Because the cost of repairing or replacing a vending machine is higher than the value of the goods it contains,damage to vending machines results in considerable loss to manufacturers and owners. Security measures such as surveillance cameras that produce reports of robberies after they have occurred cannot solve these problems. In this study an attempt is made to construct a new predictive crime-prevention system by applying the MT system to video images to classify and predict behavior.