Technology in detecting potential drop out tuberculosis (TB) in Cut Meutia hospital and Health Office plays a great role and has been very important. This is seen from the increasing number of patients who could not be cured succesfully and who do not care about TB which will have fatal consequences on their health. In addition, the main cause of the increase in the number of potential drop out TB patients is because of the lack of awareness of the community, especially the middle economic level family of the danger of TB disease as seen from the irregular treatment that they have and the continued smoking habit. In this study, an expert system was used to diagnose patients with potential Drop Out tuberculosis who were then diagnosed into the cluster of each TB patient using the K-Means algorithm. The system implementation in the expert system is that the initial symptoms include the question of whether the patient has cough with phlegm for 2-3 weeks or more (yes), has the patient been treated with TB drugs less than 1 month (no), experienced no appetite and nausea. From the results of these symptoms, there are diagnoses of New Patients, Pulmonary BTA (-) / Ro (+), with sub-acute level having moderate severity and duration, the severity can reduce the health status of the patient, the patient is eventually expected to recover and totally recovered the disease does not develop into a chronic disease. The results of this expert system would be entered into the K-Means clustering. The test results of the k-means clustering algorithm with K = 3 (C1, C2, C3). with initial centroid values of m1: C1, 5, 5, 5, 5, 5, 5 and m_2: C2, 3, 3, 3, 3, 3, with patient p1 with the value of each cluster (C1) = 6.928, ( C2) = 2.828, C3 = (4). For the closest cluster value is C2, then the BCV (Between Cluster Variation) calculation value is 19,596, and the WCV (Within cluster Variation) value is 144. Then the ratio value is 0.136. The result of the iteration -3 can be stopped because it does not experience the movement of the clusters and the clusters have been optimal. The results of this system can classify patients for each village and sub-district area so that the Hospital officials and the Health Office can directly monitor potential drop out TB patients and can facilitate the Head of Office/region in handling clustered TB patients using K-Means. Furthermore, in the coming years, it can be used as a tool in taking preventive measures.