Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Multiple testing for outlier detection in functional data
(Submitted on 13 Dec 2017)
Abstract: We propose a novel procedure for outlier detection in functional data, in a semi-supervised framework. As the data is functional, we consider the coefficients obtained after projecting the observations onto orthonormal bases (wavelet, PCA). A multiple testing procedure based on the two-sample test is defined in order to highlight the levels of the coefficients on which the outliers appear as significantly different to the normal data. The selected coefficients are then called features for the outlier detection, on which we compute the Local Outlier Factor to highlight the outliers. This procedure to select the features is applied on simulated data that mimic the behaviour of space telemetries, and compared with existing dimension reduction techniques.
Submission history
From: Clementine Barreyre [view email][v1] Wed, 13 Dec 2017 14:07:55 GMT (516kb,D)
Link back to: arXiv, form interface, contact.