Transmission line faults can cause both power loss and failure. To mitigate the effects of such faults, the energy supply must quickly identify and remove faults, as well as ensure grid restoration after a failure occurs. As a result, it is critical to design a system capable of quickly and reliably identifying and eliminating errors. The level of fault identification accuracy is an important indicator for ensuring the reliability of main equipment in power systems, such as generators and transformers. This paper proposes a two-phase approach for identifying faults in transmission systems. In the first phase, unsupervised learning techniques like K-Mean clustering are used to assign labels to datasets for transmission line fault classification. During the second phase, four machine learning techniques called logistic regression (LR), decision tree classifier (DTC), random forest classifier (RFC), and XGBoost Classifier (XGB) are employed to identify faults. Applications are validated on fault detection datasets. The tested approach provides an efficient model for fault detection and classification in transmission lines, as well as a productive framework for fault detection prediction based on machine learning and ensemble learning methods. The experimental simulation results from this study show an accuracy of 83.6% for LR, 99% for DTC, 99% for RFC, and 99.9% for XGB, LightGBM, and CatBoost at 0.99995%. The paper's findings demonstrate the effectiveness of machine learning and ensemble learning techniques in accurately identifying and classifying transmission line faults at competitive performance indices.