RESUMO
The electrical energy supply relies on the satisfactory operation of insulators. The ultrasound recorded from insulators in different conditions has a time series output, which can be used to classify faulty insulators. The random convolutional kernel transform (Rocket) algorithms use convolutional filters to extract various features from the time series data. This paper proposes a combination of Rocket algorithms, machine learning classifiers, and empirical mode decomposition (EMD) methods, such as complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), empirical wavelet transform (EWT), and variational mode decomposition (VMD). The results show that the EMD methods, combined with MiniRocket, significantly improve the accuracy of logistic regression in insulator fault diagnosis. The proposed strategy achieves an accuracy of 0.992 using CEEMDAN, 0.995 with EWT, and 0.980 with VMD. These results highlight the potential of incorporating EMD methods in insulator failure detection models to enhance the safety and dependability of power systems.
RESUMO
Predicting attachment styles using AI algorithms remains relatively unexplored in scientific literature. This study addresses this gap by employing EEG data to evaluate the effectiveness of ROCKET-driven features versus classic features, both analyzed using the XGBoost machine learning algorithm, for classifying 'secure' or 'insecure' attachment styles.Participants, fourth-year engineering students aged 20-35, first completed the ECR-R questionnaire. A subset then underwent EEG sessions while performing the Arrow Flanker Task, receiving success or failure feedback for each trial.Our findings reveal the effectiveness of both feature sets. The dataset with ROCKET-derived features demonstrated an 88.41% True Positive Rate (TPR) in classifying 'insecure' attachment styles, compared to the classic features dataset, which achieved a notable TPR as well. Visual representations further support ROCKET-derived features' proficiency in identifying insecure attachment tendencies, while the classic features exhibited limitations in classification accuracy. Although the ROCKET-derived features exhibited higher TPR, the classic features also presented a substantial predictive ability.In conclusion, this study advances the integration of AI in psychological assessments, emphasizing the significance of feature selection for specific datasets and applications. While both feature sets effectively classified EEG-based attachment styles, the ROCKET-derived features demonstrated a superior performance across multiple metrics, making them the preferred choice for this study.