Studying the Effect of Foeniculum Vulgare Mill and Ziziphora Clinopodioides Lam. Extracts on the Growth of Aspergillus Flavus Mold in Tomato Paste and Predicting the Data Obtained Using Artificial Neural Networks

The antifungal activity of Foeniculum Vulgare Mill and Ziziphora clinopodioides Lam. extracts against Aspergillus flavus in tomato paste containing different percentages of the extracts was tested. To this end, Foeniculum Vulgare Mill and Ziziphora clinopodioides Lam. extracts with different concentrations of 0.5, 1 and 2% were prepared and studied during different storage times(35 days).The effect of extracts of Foeniculum Vulgare Mill and Ziziphora Clinopodioides Lam with different concentrations was investigated alone in the environment (in vitro). By injecting 0.1 ml of mold in Sabouraoud dextrose agar broth culture medium, then placing it in an incubator temperature of 25°C ± 0.5, it was kept for 5 weeks (35 days), and one culture was done every week in order for the activity mold to be investigated in different concentrations of extracts.The results of antifungal activity of different levels of the extracts indicated that treatments 3 (containing 2% Foeniculum Vulgare Mill extract) and 4 (containing 3% Foeniculum Vulgare Mill extract) were resistant to the growth of Aspergillus flavus mold mycelium until the end of storage period.Generally, it can be concluded that using 2 or 3% Foeniculum Vulgare Mill extract as a natural preservative in tomato paste has a desirable antifungal activity. Artificial neural network was used to validate and evaluate the results of the experiments in predicting the data of Aspergillus flavus mold growth in tomato paste.In the present study, two hidden layers with 30 neurons were used. The network had two inputs including extract concentration and storage time, and the growth of Aspergillus flavus mold was considered as the target.Evaluation parameters such as correlation coefficient, mean squared error and maximum error showed very good results with values ​​of 0.9993, 0.10934 and 0.13538. The lower the error and the closer the correlation coefficient to 1, the better the prediction is.