Unsupervised statistical method for multivariate detection of atypical curves

摘要

The invention relates to an unsupervised statistical method for detecting atypical multivariate curves, on the basis of previously collected data, resulting from measurements carried out by a plurality of sensors, comprising: - a processing step 100 in which for a given sensor and for a reference curve, a similarity/dissimilarity index representative of the distance between said reference curve and each of the other curves resulting from this sensor, - first iteration 200, in which the processing step 100 is repeated for each curve, iteratively so as to obtain a similarity/dissimilarity index for each curve relative to the other curves, - second iteration 300, so as to carry out steps 100 to 200 with the other sensors, so as to obtain, for each of these sensors, a table of similarity/dissimilarity indices, - an identification step by carrying out a multivariate statistical treatment on similarity/dissimilarity indices resulting from the second iteration step. Figure for abstract: Fig. 1