Author(s):
1. Milica Karadžić, University of Novi Sad, Faculty of Technology, Serbia, Serbia
2. Lidija Jevrić, University of Novi Sad, Faculty of Technology, Serbia, Serbia
3. Sanja Podunavac Kuzmanović, University of Novi Sad, Faculty of Technology, Serbia, Serbia
4. Strahinja Kovačević, University of Novi Sad, Faculty of Technology, Serbia, Serbia
5. Aleksandra Tepić Horecki, University of Novi Sad, Faculty of Technology, Serbia, Serbia
6. Zdravko Šumić, University of Novi Sad, Faculty of Technology, Serbia, Serbia
7. Senka Vidović, University of Novi Sad, Faculty of Technology, Serbia, Serbia
8. Žarko Ilin, Faculty of Agriculture, University of Novi Sad, Novi Sad, Serbia, Serbia

Abstract:
The aim of this study was to predict the antioxidant activity of selected lettuce cultivars (Lactuca sativa L.) using non-linear approach. The data set contained 31 samples and 6 phytochemicals contents were used for chemometric modeling. Using partial least square (PLS) method the most important phytochemicals were selected and used as inputs in the further step. The prediction was achieved by using self-training artificial neural networks (ST ANN) approach. By using sum of ranking differences (SRD) method the generated ANN models were ranked. The aim was to distinguish the most consistent ANN models. The phytochemicals most affecting antioxidant activity are total phenols and chlorophyll a+b content. Global sensitivity analysis (GSA) gave insight into each phytochemical contribution within generated ANNs. According to the standard statistical parameters, graphical methods and sum of ranking differences approach the established ANNs could be used for the prediction of the antioxidant activity.

Key words:
antioxidant activity, global sensitivity analysis, phytochemicals, sum of ranking differences

Thematic field:
Food Technology

Date of abstract submission:
19.10.2016.

Conference:
XI Savjetovanje hemičara, tehnologa i ekologa Republike Srpske

Abstract