· ·

Version 5.1.0 Release!

· ·

KavaLytics™ V 5.1.0 release is a major update to the KavaLytics™ Predictive Models.

This update incorporated 49 unique samples from Fiji, Vanuatu, Solomon Islands and Hawaii. A high level summary of the updates for Version 5.1.0 are the following:

  • New Beverage Grade vs. Non Beverage Grade Classifier - 95% Accuracy
  • New Flavokavain Level warning for samples predicted above 0.25% FKA+B - 92.2% Accuracy
  • Increased predictive accuracy on stems and peelings
  • Increased predictive accuracy on previously anomalous samples

The "Noble" or "Non-Noble" Classifier has been integrated into a new "Beverage Grade" or "Non-Beverage Grade" Classifier to match the current regional, and soon to be international, WHO/FAO Kava Codex. "Beverage Grade" samples have Noble Chemotypes, and are comprised of acceptable plant parts and whereas “Non-Beverage Grade” samples do not meet at least one of these criteria. This classifier correctly classifies 530 out of 558 samples in the training set (95%) and is intended to be used alongside the Kavalactone quantification model, Flavokavain qualitative classifier, and moisture content model to make a final determination of the quality of the material.

The Flavokavain quantification model has improved significantly thanks to high flavokavain samples from the Solomon Islands, Vanuatu, and Hawaii. The quantification model will remain a beta feature until a 90% level of accuracy is achieved. However, a new Qualitative classifier has been built based upon the data in this model and is able to achieve 92.2% accuracy at flagging samples above 0.25% Flavokavain A+B and will now be a new powerful standard feature of all future KavaLytics™ models.

The high Flavokavain samples from Hawaii were samples from Noble Hiwa Plants that the previous model had flagged as Non-Noble. These samples were sent for further testing and it was found that they indeed had Non-Noble levels of Flavokavains, despite their Noble genetics. These plants were grown under stressful environmental conditions which in turn caused them to have Non-Noble levels of Flavokavains:

High Flavokavain Hawaiian Kava Black Hiwa

Similarly, a sample from the Solomon Islands was flagged by our previous model as Non-Noble, despite its Noble 426 Chemotype and strikingly high potency of 11.4%. Upon further analysis, it was found that this sample had extremely high Flavokavain levels and should indeed not be considered Noble Kava:
High Flavokavain Solomon Kava
The above examples demonstrate the need for kava testing even when it is known to come from Noble cultivars or appear by its Kavalactone chemotype or acetone test to be Noble. It also shows that our predictive algorithms are extremely advanced and can detect even the most unsuspecting problematic kava material that passes other quality testing methods.

We are tirelessly working on improving these models and providing our customers with the most advanced Kava Quality Control and Traceability system the world has ever known. With each scan, the system becomes more and more accurate at its predictions and opportunities for new features jump out at us as we toll through the data.  Thank you to our loyal customers, its our absolute pleasure to continue to develop these models for the mutual benefit of the Kava Industry.