Transmission lines are fundamental components of the electric power system, demanding special attention from the protection system due to the vulnerability of these lines. This paper presents a method for fault location in transmission lines using data for a single terminal without requiring explicit feature engineering by a domain expert. The fault location task provides an approximate position of the point of the line where the failure occurred, serving as information to the operators to dispatch a maintenance staff to this location to reclose the transmission line with better reliability and safety. In our method, we extract two post-fault cycles of the three-phase current and voltage signals to serve as input to a model based on the LSTM algorithm. We defined the model's architecture with empirical experiments searching for the best structure to estimate the fault distance. For this purpose, we used a dataset with diversified failure events, also available to the scientific community. The results demonstrate the effectiveness of the proposed method with a mean error of 0.1309 km +- 0.4897 km, representing 0.0316% +- 0.1183% of the transmission line extension.
Faramarz Farhangian, Leandro Augusto Ensina, George D C Cavalcanti, Rafael M. O. Cruz. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
The Brazilian Higher Education Census has revealed that the dropout rate among higher education students in Brazil exceeds 50% starting from the fifth year. This high rate results in several problems related to the wastage of resources invested by both the society and the students. Therefore, universities need to develop strategies to prevent student dropout and minimize these problems. However, predicting student dropout involves detecting patterns and predicting them over a large amount of data collected yearly from thousands of students. Given the scale and volume of data involved in dropout prediction, machine learning emerges as a powerful technique to automate the identification of these students. The objective of this paper is to identify students who are prone to dropping out based on the academic history of Bachelor’s Degree in Computer Science students at an unpaid public university in Brazil. We engineered four datasets based on the semester in which the students are enrolled. These datasets are designed to simulate the academic scenario and individual characteristics of the students available up to the prediction moment. Besides, we propose three feature models to identify the best scenario. Our method could identify the students most likely to drop out and the main features that contributed to the respective decision. Using only the information from the disciplines taken by the students proved to be the best feature model. When using these features with Gradient-Boosting, the F1-Score performance ranged between 69% and 85%, depending on the dataset.
This paper introduces an approach for discovering denial constraints (DCs) to identify faults in transmission lines. However, the considerable volume of data in the studied scenario makes traditional DC discovery impractical due to lengthy execution times. We propose an alternative DC discovery approach that uses streaming windows to address this issue. Our experiments demonstrate that the DCs identified in pre-fault windows differ significantly from those in post-fault windows. This valuable insight enables us to detect faults autonomously, eliminating the need for human intervention (i.e., an unsupervised method). The experimental evaluation featuring diverse fault events reveals that our approach achieves fault detection with remarkable 100% accuracy. This paper introduces an approach for discovering denial constraints (DCs) to identify faults in transmission lines. However, the considerable volume of data in the studied scenario makes traditional DC discovery impractical due to lengthy execution times. We propose an alternative DC discovery approach that uses streaming windows to address this issue. Our experiments demonstrate that the DCs identified in pre-fault windows differ significantly from those in post-fault windows. This valuable insight enables us to detect faults autonomously, eliminating the need for human intervention (i.e., an unsupervised method). The experimental evaluation featuring diverse fault events reveals that our approach achieves fault detection with remarkable 100% accuracy.
The rapid spread of information via social media has made text-based fake news detection critically important due to its societal impact. This paper presents a novel detection method called Dynamic Representation and Ensemble Selection (DRES) for identifying fake news based solely on text. DRES leverages instance hardness measures to estimate the classification difficulty for each news article across multiple textual feature representations. By dynamically selecting the textual representation and the most competent ensemble of classifiers for each instance, DRES significantly enhances prediction accuracy. Extensive experiments show that DRES achieves notable improvements over state-of-the-art methods, confirming the effectiveness of representation selection based on instance hardness and dynamic ensemble selection in boosting performance. Codes and data are available at: https://github.com/FFarhangian/FakeNewsDetection_DRES