With increasing frequency, teenagers and young adults are being diagnosed with such mental and behavioral health issues as depression, anxiety, mood disorders and addiction. In many cases, these conditions are not detected in a timely fashion-and sometimes not detected at all-thereby preventing access to needed treatment options by the patients affected. These conditions are often concomitant with sharply declining or widely fluctuating academic performance in school. Our long-term goal is to promote an early detection of mental and behavioral health issues in high school students through a statistical analysis of their academic performance. Here we discuss unsupervised learning methods seeking to classify students into three clusters of stable, monitor and critical students. A key goal of our work is to produce clusters whose size is consistent with statistics for adolescent mental health posted periodically by the US Centers for Disease Control and Prevention (CDC) and the National Alliance on Mental Illness (NAMI). Our analyses are based on a data set of academic performance data involving over 500 students in mid-sized high schools in the state of Illinois.