Obesity is an endemic in most part of the developed word. As the increased occurrence of many serious comorbodity (stroke, IHD, NIDDM) is casually linked to obesity, it represents a huge risk from the public health point of view. Focus should be placed on early identification of risked individuals, which calls for effective screening methods. This is largely hindered by the fact that current diagnostic methods are of either poor predictive value (anthropometric indices etc.) or unavailable for large-scale screenings (BIA, DXA etc.) Our idea to resolve this problem is based on the fact that obesity has a marked effect on many of the routinely used laboratory parameters. An obvious example is the serum level of various blood lipids: hyperlipidemia, hypercholesteremia are often observed in obese people. On the grounds of this fact, we hypothesized that obesity can be predicted using only routine laboratory parameters. Furthermore, we presumed that those are the best parameters to predict obesity (and obesity-associated risk), that best separate manifestly obese and healthy people. Our research aimed to investigate this possibility on adolescent population, as they are the most important from the public health point of view. To that end, we performed a cross-sectional clinical study that included the observation of n=148 male children (aged 12-16 year), consisting of healthy volunteers from four Hungarian secondary schools and obese patients treated with E66.9 “Obesity, unspecified” diagnosis. Observation included the recording of 27 laboratory parameter from a fasting blood sample. To investigate how well obese and healthy children can be separated based solely on their laboratory results, we employed a state-of-the-art classification tool, Support Vector Machines (SVM). Results were compared with the more classical approach of multivariate logistic regression. SVM’s major drawback is its “blackbox” nature; however, we concluded that its performance is excellent, superior to logistic regression.