Application of artificial neural network for tracing the geographical origins of coffee bean in northern areas of Thailand using near infrared spectroscopy

Abstract

The aim of this research study was to investigate the difference among coffee bean from different plantation areas in the northern of Thailand. Near infrared (NIR) spectra were recorded from Arabica coffee samples which were collected from Chiang Mai, Lampang and Mae Hong Son provinces in Thailand. In addition, color parameters and moisture content were analyzed. The data were exploratorily analyzed based on the uses of principal component analysis (PCA) and an artificial neural network (ANN) called self-organizing map (SOM). To identify the significant parameters of the spectroscopic data, a variable selection called self-organizing map discrimination index (SOMDI) was applied. As a result, SOM could overcome the PCA technique where the samples from the three different origins could be separated. Additionally, based on the SOMDI results, the coffee samples from Chiang Mai could be well discriminated using the NIR spectral regions of 880-1182, 1254-1326, 1896-2180 and 2260-2498 nm. This research demonstrated that using NIR spectroscopy coupled with the ANN algorithm allowed an efficient tracing method to differentiate the coffee bean samples in the northern of Thailand.