Escrever uma avaliação PREreview

Time series data, characterized by temporal dependencies, seasonality, and noise, are prevalent in domains such as healthcare, finance, energy, and transportation. Effective clustering of time series enables the discovery of patterns, supports forecasting, and facilitates data-driven decision-making. This paper provides a comprehensive review of time series clustering techniques, including conventional methods (e.g., k-means, hierarchical, and fuzzy clustering), similarity-based approaches (e.g., Dynamic Time Warping), feature-based methods, and deep learning models (e.g., autoencoders, convolutional and recurrent neural networks). The review analyzes tasks, application domains, performance outcomes, and key limitations, highlighting common challenges such as computational complexity, sensitivity to noise, and scalability issues. A particular focus is given to transport-related time series, including traffic flow, travel time, and congestion patterns, demonstrating how clustering can support traffic state classification, anomaly detection, and infrastructure planning. The analysis reveals a trade-off between accuracy, interpretability, and computational efficiency, emphasizing the need for scalable, robust, and domain-aware clustering frameworks. Finally, practical directions for future research are discussed, including lightweight hybrid approaches and transport-specific feature engineering to enhance clustering performance in real-world applications.