In diagnosing mental illness takes a relatively long time is one week to one month and could only be carried out by experts which is psychological doctors, because it must diagnosing accordance with existed procedures. Based on the above the problem can be formulated how to diagnose psychiatric illness by applying neural network Learning Vector Quantuzation 2 (LVQ 2). LVQ2 is the development of basic LVQ. This study aims to help doctors diagnose phychiatric  illness by applying neural network Learning Vector Quantization 2 and can distinguish between types of mental illness. In this research using the data input 14 symptoms and 4 psychiatric diseases as output is used as the target of Schizophrenia, Organic Mental Disorders, Mental disorders and behavior due to substance users, and feeling the atmosphere Disorders (affective or mood disorders). Based on test results using 132 training data and 30 test data and parameter with a value of learning rate = 0.025, a reduction in the minimum learning rate = 0.1 learning rate = 0:01, and window = 0.4 which is done LVQ2 test result accuracy is as high as 90%. Thus LVQ2 can be applied to the classification of mental illness.