Ph.D. Dissertation:

Application of Neural Network in Modeling the effect of adding green tea٫ white tea and ginger extracts on physicochemical٫ chemical and sensory properties Sponge cake

In recent years, the use of plant sources with bioactive compounds such as pigments and antioxidants in food has become widespread .Research has shown that these resources play an important role in improving quality properties, enhancing nutritional value and increasing the shelf life of food products such as cereals and dairy products .So, in the present study, single levels of green tea extract (zero to 2 %), white tea (zero to 2 %) and ginger (zero to 1/5 %) were used in the sponge cake formulation and chemical, physicochemical and sensorial tests were performed on different treatments. To optimize the results, the response surface methodology (RSM) was used. In order to predict the experiments results, the multilayer perceptron neural network with two inputs (extract concentration, storage time) and one output using MATLAB R2113a software was used. The learning model of Levenberg-Marquardt, the learning cycle 1111, and correlation coefficient (R2), mean square error (MSE), Root mean squared error (RMSE) and Mean absolute error (MAE), Mean absolute percentage error (MAPE), or Mean absolute percentage deviation (MAPD) were created and evaluated for network. According to the results, with increasing green tea, white tea and ginger extracts, the fat and protein content, total phenol, pH value, guminess and chewiness increased in the treatments .In contrast, moisture content, acidity, peroxide value, cohessiveness, springness, resilience and sensory scores decreased. The level of mold and yeast contamination of the samples on the first, seventh and fourteenth days was equal to logarithm 1 and the highest level of contamination was observed on the twenty-first day. All three extracts of green tea, white tea and ginger extract had significant effects on the physicochemical, physical and sensory properties of sponge cake samples. It can be said that the optimal sample lasts until the fourteenth day without microbial contaminatio. The results of the RSM showed that the low levels of the extracts gave better results in the studied parameters except peroxide value .The artificial neural network (ANN) model had a high ability to predict most of the parameters studied in the present study and had a good fit with experimental data .The results showed that the ANN was an effective tool for predicting the results of qualitative experiments on sponge cake samples.

Keywords: Key words: Tea, Ginger, Artificial Neural network, Sponge cake.