Deep Learning Techniques to Improve the Performance of Olive Oil Classification.

Area of Science:

• Analytical Chemistry
• Food Science
• Artificial Intelligence

Background:

• Traditional olive oil assessment relies on sensory panel tests.
• Novel strategies using Gas Chromatography (GC), mass spectrometry (MS), and ion mobility spectrometry (IMS) with chemometrics are of interest for olive oil classification.
• Accurate classification is crucial for preventing fraud and ensuring consumer safety.

Purpose of the Study:

• To combine chemical techniques (GC-IMS) and Deep Learning (DL) for automatic classification of olive oil samples.
• To classify olive oil into three categories: extra virgin olive oil (EVOO), virgin olive oil (VOO), and lampante olive oil (LOO).
• To develop a robust classification model using data from two harvests (2014-2015 and 2015-2016).

Main Methods:

• Utilized Gas Chromatography-Ion Mobility Spectrometry (GC-IMS) to obtain spectral fingerprints from 701 olive oil samples.
• Selected specific olive oil markers from the spectral data.
• Developed and configured a Deep Learning model tailored to the specific data characteristics.

Main Results:

• The Deep Learning model successfully classified olive oil samples into EVOO, VOO, and LOO categories.
• Achieved higher success rates compared to previously reported methods.
• Demonstrated the effectiveness of combining instrumental chemical techniques with DL for oil classification.

Conclusions:

• Deep learning approaches applied to data from chemical instrumental techniques provide an effective method for classifying olive oil samples.
• The developed model offers improved accuracy for olive oil quality assessment.
• This approach supports robust, automated classification for quality control and fraud prevention.