Bearings are critical components in rotatory machineries. Unexpected failure in the bearings causes huge impact in industries. Predicting the remaining useful life time of bearing helps in replacing them without any delay and allowing maintenance. Numerous research works have done in predicting the life time of bearings and many of the works were based on regression model. In the present work, classification approach was carried out and a predictive model was built to assess the remaining useful life time of bearings. Vibration signals will be acquired on continuous basis from bearings running in the experimental set-up. Vibration signals were used in the present study which was operated at rated speed and load conditions. A brand new bearing was taken for the experiment and the signals were acquired on daily basis till it fails naturally, meaning run to failure tests was carried out. In this paper, decision tree is used for feature selection and comparative study of Naïve Bayes and Bayes net classifiers in predicting remaining useful life time was carried out. The result shows that Bayes net classifier gives 89.64% classification accuracy with 5 features whereas Naïve Bayes classifier yields 77.16% accuracy. © Research India Publications.