Rolling element bearing is widely used in industrial machineries to support load and fault in rolling element bearing can lead to catastrophic failure to the entire plant. Therefore, a reliable bearing problem diagnosis technique is essential for preventing machinery malfunction. In modern days, using different machine learning (ML) techniques, online monitoring of health status of ball bearing can be easily performed. Among all the ML techniques, K-Nearest Neighbor (KNN) is one of the most favored techniques in fault diagnosis in rotating machinery, as it is user-friendly and robust to the effect of noise. In the present work, experiments are conducted using a bearing test rig for three different fault conditions (outer, inner, and ball fault) of a ball bearing. The useful features are extracted from the vibration response (system with healthy as well as faulty bearing) to train the KNN model. An accuracy of 98% is achieved in detecting the condition of bearing that reveals the effectiveness of the proposed model. This trained model can further be utilized for determining the condition of bearings at different operating conditions.