Bearing degradation is one of the primary causes of failure in rotating machinery, and vibration-based analysis is one of the most effective techniques for diagnosing and monitoring bearing conditions. This study presents a vibration data-driven approach for assessing the bearing health condition and predicting remaining useful life (RUL) during the final stage of degradation. Time-domain features are extracted and used to classify the bearing condition, then to predict RUL during the last degradation stage. The performance of the classification and RUL prediction using support vector machine (SVM), support vector regression (SVR), and adaptive neuro-fuzzy inference system (ANFIS) demonstrates the effectiveness of the proposed method and emphasis the relevance of the vibration-based techniques for predictive maintenance applications.