Multivariate functional outlier detection using the fast massive unsupervised outlier detection indices

We present definitions and properties of the fast massive unsupervised outlier detection (FastMUOD) indices, used for (OD) in functional data. FastMUOD detects outliers by computing, each curve, an amplitude, magnitude, shape index meant to target corresponding types outliers. Some methods adapting multivariate data are then proposed. These include applying on components using random projections. Moreover, these techniques tested various simulated real datasets. Compared with state art OD, use projections showed most effective results similar, some cases improved, OD performance. Based proportion that flag function as outlier, we propose a new graphical tool, magnitude-shape-amplitude (MSA) plot, useful visualizing amplitude outlyingness