UNSUPERVISED STATISTICAL METHOD FOR MULTIVARIATE IDENTIFICATION OF ATYPICAL SENSORS

摘要

The invention relates to a method for identifying atypical sensors measuring characteristics of individuals, comprising: - a step of collecting 50 curves of a characteristic of individuals, the curves being measured by each sensor; - a processing step 100 wherein, for a given sensor and for a reference curve, an index of dissimilarity between said curve and each of the other curves of the sensor is calculated; - a first iteration step 200, wherein the processing step 100 is iteratively repeated for each curve resulting from the same sensor to obtain a dissimilarity index for each curve; - a second iteration step 300, wherein steps 100 to 200 are carried out with the other sensors to obtain a table of dissimilarity indices; - a step of calculating 400 an atypicality index for each individual from multivariate statistical processing of the tables; - a step of identifying 600 atypical individuals; - a step of identifying 700 atypical sensors.