Classification of olive oils using chromatography, principal component analysis and artificial neural network modelling

Gümüş Z. P. , Ertaş H. , Yasar E. , Gümüş Ö.

JOURNAL OF FOOD MEASUREMENT AND CHARACTERIZATION, cilt.12, ss.1325-1333, 2018 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 12
  • Basım Tarihi: 2018
  • Doi Numarası: 10.1007/s11694-018-9746-z
  • Sayfa Sayıları: ss.1325-1333


Classification and addressing, and geographical origin of different olive oils is of great importance due to their differentiation in quality, and for commercial concerns. In this study, quantification of sterols, fatty acids, and triacylglycerol composition of forty-nine olive oils collected from six different locations of western part of Turkey (Izmir, Manisa, Aydm, Mugla, Bursa, and Edremit Bay) were performed by using chromatographic methods. Data for those olive oil samples were compiled, and classified with the artificial neural network (ANN) modelling and principal components analysis (PCA). The analytical results included resourceful information about determining geographical origin and traceability of olive oil in Turkey by using ANN and PCA. The ANN model for sterol composition showed the highest accuracy with 85.71%. The FAME and TAG profiles followed this with 83.67 and 81.63% accuracy respectively. However, Izmir and Manisa regions have poor sensitivity values with all ANN models since they are geographically very close to each other. Furthermore, the PCA results of the sterol composition have provided separation and clustering between locations. beta-sitosterol, campesterol, stigmasterol and 24-metilen cholesterol have an important role in determining the separation of the locations of origin. While separation of the Bursa location has been under the pressure of FAME composition, the TAGs have been effective on the clustering of the Aydm and Edremit Bay. In conclusion, the geographical authentication of Turkish olive oils can be done with high accuracy by using ANN and PCA.