Machine Learning Pins Down Where Edible Seaweed Comes From

Sargassum fusiforme, a high-value edible seaweed harvested across East Asia, now has a new line of defense against mislabeling. A research team reports that combining near-infrared spectroscopy, elemental profiles, and stable isotope ratios with machine learning can identify whether samples of the seaweed came from China, Japan, or Korea. The study, published in Chemical and Biological Technologies in Agriculture, systematically tested each data source alone and in combination to see which strategy best authenticated the geographical origin of the species.

Geographical authentication matters because origin is closely tied to the quality and traceability of this seaweed, which commands high market value. Consumers and regulators alike need reliable ways to confirm that a product labeled as coming from a particular country genuinely does. The researchers set out to determine whether chemical fingerprints embedded in the seaweed itself could serve that purpose, using instruments that measure different aspects of its composition.

The team compared three machine learning algorithms: linear discriminant analysis, support vector machine, and random forest. They worked with 53 samples of S. fusiforme drawn from the three countries. Near-infrared spectroscopy captures information about organic molecules in the tissue, elemental profiling records the concentrations of individual chemical elements, and stable isotope ratios reflect isotopic signatures that can vary with local growing conditions. Each of these data types was tested on its own, in pairs, and all together.

Single-feature models, meaning models built on just one of the three data sources, showed limited performance. Most of these models achieved prediction accuracies below 70 percent, which is not sufficient for dependable origin verification. The result suggests that no single chemical fingerprint, whether spectral, elemental, or isotopic, carries enough country-specific signal on its own to separate the three origins reliably.

Pairing data sources changed the picture considerably. Two-way combinations, such as elemental profiles combined with near-infrared spectroscopy or stable isotopes combined with elements, improved both accuracy and stability. These paired models reached 81.25 percent prediction accuracy when built with either support vector machine or random forest algorithms. The improvement indicates that different data sources capture complementary information about the seaweed’s origin.

The strongest performance came from using all three data sources at once. When stable isotope ratios, elemental profiles, and near-infrared spectra were combined, the random forest model achieved the highest prediction accuracy of the study, 87.5 percent. Random forest is an ensemble method that builds many decision trees and aggregates their votes, which appears well suited to integrating the heterogeneous information from these different measurement types.

Beyond raw accuracy, the researchers examined which variables drove the classification. Feature importance analysis identified copper, arsenic, and the nitrogen isotope ratio δ15N as the three most influential variables. Copper carried an importance value of 0.1880, arsenic 0.1862, and δ15N 0.1751. These values indicate that trace element concentrations and nitrogen isotopic composition contributed more to distinguishing origins than most other measured features, likely reflecting differences in local environmental and growing conditions across the three countries.

The authors conclude that integrating multiple data sources with random forest offers a promising strategy for geographical authentication of S. fusiforme. Because the study relied on 53 samples and three countries, the reported accuracies reflect that specific dataset and scope. Even so, the finding that combined chemical fingerprints outperform any single measurement type offers a practical template for traceability efforts in high-value seaweed supply chains, where confirming origin protects both product quality and market integrity.

Subject of Research: Machine learning combined with spectroscopy, elemental profiles, and stable isotope ratios to authenticate the geographical origin of an edible seaweed.

Article Title: Combined use of near-infrared spectroscopy, elemental profiles, and stable isotope ratios with machine learning for geographical origin authentication of Sargassum fusiforme

Article References: Ji, X., He, Q., Wei, S., Chen, Y., Liang, Y., Yu, J., Yang, Y., & Liao, Z. (2026). Combined use of near-infrared spectroscopy, elemental profiles, and stable isotope ratios with machine learning for geographical origin authentication of Sargassum fusiforme. Chemical and Biological Technologies in Agriculture. https://doi.org/10.1186/s40538-026-01115-7

Image Credits: AI Generated

DOI: 10.1186/s40538-026-01115-7

Keywords: Sargassum fusiforme, geographical origin authentication, near-infrared spectroscopy, elemental profiles, stable isotope ratios, machine learning, random forest, support vector machine, linear discriminant analysis, food traceability, seaweed, feature importance