公开(公告)号：US20200065699A1

公开(公告)日：2020-02-27

申请号：US16462611

申请日：2017-01-19

Applicant: SHIMADZU CORPORATION

Inventor： Yusuke TAMAI ,  Shigeki KAJIHARA ,  Shin FUJITA ,  Ryota AISU

IPC: G06N20/00

Abstract: This analytical data analysis method uses machine learning of analysis result data (31) measured by an analyzer (1), and includes generating simulated data (32) in which a data variation has been added to the analysis result data (31) within a range that does not affect identification, performing the machine learning using the generated simulated data (32), and performing discrimination using a discrimination criterion (23b) obtained through the machine learning.

公开(公告)号：US20190041369A1

公开(公告)日：2019-02-07

申请号：US15773702

申请日：2015-11-05

Applicant: SHIMADZU CORPORATION

Inventor： Ryota AISU

IPC: G01N30/86 ,  H01J49/26 ,  G01N30/72

Abstract: A compound-weight time-series data calculator creates a compound spectrum matrix based on standard mass spectra of various compounds stored in a standard mass spectrum library. Then, the calculator determines, in the form of a linear regression model, the relationship of the matrix, a measured vector based on the data acquired at one measurement time point, and a compound weight vector at the measurement time point, and estimates the unknown weight vector by a minimum norm estimation in which a regularizer is introduced. A compound-weight time-series graph creator creates a compound-weight time-series graph showing a temporal change in weight for each target compound, based on the compound weight vector obtained at the measurement time points within a specified time range. A peak which appears on this graph is used for the quantitative determination of the compound or determination on the presence or absence of the compound.