Identifying key parameters to maximize lavender essential oil components during supercritical extraction using machine learning models

The purpose of our work is to quantitatively compare the impact plant rearing and extraction parameters have on the composition of Lavandula angustifolia essential oil (EO) using machine learning techniques. It is critical to assess and rigorously compare the impact of process conditions on the final product and select key process parameters to optimize yield and composition of lavender EO. In our work, lavender EO obtained through supercritical fluid extraction was studied and the relative importance of the extraction parameters, such as pressure (100–200 bar) and temperature (40–60 °C), as well as conditions of rearing and plant processing such as elapsed time since harvesting (1–20 months), place of origin, and environmental effects (average temperature, pre-flowering precipitation) on the yield and composition were ranked using machine learning models. It was found that rearing conditions, especially precipitation, most heavily determine EO composition. This is based on the LMG indices of linalool and 4-terpineol, which were 0.38 and 0.25. However, extraction temperature significantly affects yield, with an LMG index of 0.13. The identified key parameters can be used to formulate multi-objective optimization problems to find rearing and extraction conditions for obtaining EO with desired composition. • Comparison of plant rearing, preprocessing, and extraction for oil composition. • Machine learning identifies parameters to maximize oil component yields. • Selection of key factors for multi-objective optimization of lavender oils. • User-defined composition goals guide optimization of essential oil profiles.