Analyzer, analysis method, analyzer program, and analysis learning device

An analysis method that analyzes a gaseous measurement sample on a basis of spectrum data obtained by irradiating the gaseous measurement sample with light, the analysis method comprising:

• on a basis of a total analysis value of a gaseous reference sample of which the total analysis value of multiple predetermined components are preliminarily obtained, calculating a total analysis value of the multiple components in the gaseous measurement sample from the spectrum data of the gaseous measurement sample, wherein the total analysis value of the gaseous reference sample is a quantitate property of the gaseous reference sample;
• storing data indicating a correlation, obtained using machine learning, between spectrum data of the gaseous reference sample and the total analysis value of the gaseous reference sample, wherein each spectrum of the spectrum data is a single spectrum containing contributions of all components in the gaseous reference sample;
• and
• applying the correlation to the spectrum data of the gaseous measurement sample to calculate the total analysis value of the multiple components in the gaseous measurement sample.

An analyzer that analyzes a gaseous measurement sample on a basis of spectrum data obtained by irradiating the gaseous measurement sample with continuous light having a plurality of wave numbers, the analyzer comprising:

• a total analysis value calculation part that, on a basis of a total analysis value of a gaseous reference sample of which the total analysis value of multiple predetermined components are preliminarily obtained, calculates a total analysis value of the multiple components in the gaseous measurement sample from the spectrum data of the gaseous measurement sample, wherein the total analysis value of the gaseous reference sample is a quantitate property of the gaseous reference sample and wherein the total analysis value calculation part includes
• a correlation data storage part that stores data indicating a correlation, obtained using machine learning, between spectrum data of the gaseous reference sample and the total analysis value of the gaseous reference sample, wherein each spectrum of the spectrum data is a single spectrum containing contributions of all components in the gaseous reference sample,
• a calculation main body part that applies the correlation in the correlation data storage part to the spectrum data of the gaseous measurement sample to calculate the total analysis value of the multiple components in the gaseous measurement sample, and
• a correlation calculation part that, on a basis of the spectrum data of the gaseous reference sample and the total analysis value of the gaseous reference sample, calculates the correlation from respective total analysis values and respective pieces of spectrum data of multiple gaseous reference samples to store it in the correlation data storage part, and when a total analysis value of a new gaseous reference sample and its spectrum data are acquired, takes into account the total analysis value of the new gaseous reference sample and its spectrum data to update the correlation.