Anomaly Detection in Catalog Streams

Cleaning Big Data Streams: A Systematic Literature Review

“…A total of 74% of previous studies compared their proposed method with existing methods using running time and metrics like precision, recall, accuracy, and F1 score. For instance, Ma et al [68] and Yang et al [72] compared their method with several methods, like linear regression, K-nearest neighbors, and recurrent neural networks, to validate the effectiveness of proposed method in dealing with missing values. Only 14% of studies evaluated the suggested method by using metrics without a comparison with other methods.…”

“…Two datasets, the Cyclone Wildfire Flood Earthquake Database and the Comprehensive Disaster Dataset, were used, and the results of the proposed module were compared to those of existing methods in terms of data compression and accuracy. Yang et al [72] proposed the filteringidentifying-based anomaly detection algorithm, which comprised two complementary methods, the first of which combined identifying and filtering methods, and the second of which combined data-oriented general and data-oriented specific methods for global outlier detection. Data were collected from a ground-based wide-angle camera on the third day of the experiment using a catalog stream dataset and catalog stream located dataset, and the results were compared to those achieved with drift spot, long short-term memory nonparametric dynamic thresholding, normalized feature deviation, and Wavelet in terms of F1-score.…”