Outlier Rejection in Massive Trace Element Data for Geographical Origin Identification System of Vegetables

As the result of establishment of brands for some farming districts in Japan, it is a big social issue that agricultural products of which geographical origins are camouflaged. To solve this problem, we have proposed a distributed system that identifies cultivated places of vegetables by using their trace element content as coresearch with agricultural chemists. This system records trace element content of vegetables when shipping from farms into databases which located in various farming districts. In case of a vegetable of which geographical origin is doubtful was found in markets, this system estimates real geographical origin by comparison between its trace element content and trace element content stored in databases. This system requires to measure and store trace element content in short time. Because amount of trace element data is relatively numerous, commitment of humans’ works should be avoided as possible as. On the other hand, since a lot of outliers are contained in trace element data, it is important that not to store trace element data which contains outliers to realize correct geographical origin identification in this system. In this paper, we describe a method for our geographical origin identification system to find and reject outliers in massive quantity of trace element content data.