Transmission lines are the key elements of power distribution systems, as they enable the power system to transfer electricity from one end of a country to another. It is crucial to identify the faulty lines immediately and disconnect the faulty lines to prevent the unnecessary discharge of bulk power through the faulted point. Removing faulty lines as soon as possible enables the system to restore power system stability and resume normal power flow operations. The prediction of the electrical faults and their classification is necessary to perform some control operations, such as operating a particular relay and disconnecting that particular faulty section from the power grid. Machine learning algorithms have been used to predict and classify the particular fault based on the phase voltage and current in particular lines. Case studies are performed based on the fundamental input features and the derived input features calculated from the fundamental features. The symmetrical components of voltage and current, calculated from the fault analysis, have been utilized as the derived features and employed to enhance the performance of various machine learning algorithms. It is observed that these derived input features improve the accuracy of logistic regression from 32–52%, decision tree from 89–98%, and Adaboost from 54–63%. The performance indices used for the comparison of these different machine learning algorithms are confusion matrix parameters and the classification report of each algorithm. Additionally, the study focuses on the power quality analysis of system parameters once the employed machine learning algorithms identify the faults.