Detecting and classifying outliers in big functional data

We propose two new outlier detection methods, for identifying and classifying different types of outliers in (big) functional data sets. The proposed methods are based on an existing method called Massive Unsupervised Outlier Detection (MUOD). MUOD detects classifies by computing each curve, three indices, all the concept linear regression correlation, which measure outlyingness terms shape, magnitude amplitude, relative to other curves data. ‘Semifast-MUOD’, first method, uses a sample observations while ‘Fast-MUOD’, second point-wise or $$L_1$$ median indices. classical boxplot is used separate indices from those typical observations. Performance evaluation using simulated show significant improvements compared MUOD, both computational time. that Fast-MUOD especially well suited handling big dense datasets with very small time methods. Further comparisons some recent also superior comparable accuracy apply weather, population growth, video