This paper proposes an approach to create a smart system using machine learning technique to predict the maintenance alert for industrial motors before it breaks down. Today the industrial motor user has to wait for either motor break-down or motor starts functioning below expectation and then start to look out for repair to resume the motor back to normalcy. The delay in recognizing the motor malfunction leads to unexpected downtime and higher repair costs. This research looks for proposing a predictive approach that perform calculations on real-time data received from motor attached sensors such as vibration, current, and temperature. The input data will be analyzed by computer using machine language technique to find anomaly in any patterns received from sensor and predict when equipment might break. The prediction will help to take timely repairs activity to avoid the motor breakdown. The five different machine learning (ML) algorithms are used in the research analysis environment: random forest (RF), support vector machine (SVM), k-nearest neighbour (KNN), both naïve Bayes (NB) and linear regression (LR). The models will be evaluated to identify which model is better at forecasting failures, therefore the trained models will be evaluated using data from operational motors. The random forest model predicts motor failure times the best among rest other models as per the test results. This data-driven predictive maintenance system minimizes downtime and maintenance costs of motor and also helps to take a decision while scheduling maintenance of the motor. The usage of Industrial IoT (IIoT), MQTT messaging, and machine learning together will certainly make a remarkable impact in industrial motor maintenance and will align with Fourth Industry revolution ideology.