Determination of turkish delight quality criteria withartificial intelligence application

The main obstacle to sustained manufacture of Turkish delight is maintaining its quality. The key components of its production are the ideal ratios of sugar, starch, and citric acid as well as the manufacturing process and expertise. This study used response surface methodology (RSM) using design of experiment (DOE) to optimize the quality of Turkish delight. After choosing 20 runs in total to make Turkish delight, quality metrics including pH, total starch, total sugar, and total moisture content were examined, followed by analysis using response surface regression model. Professionals in the Turkish delight sector graded the prepared samples according to their look, taste, odor, and flexibility. Data validation was conducted using both classical and quantum machine learning models that used classification-based model. The results showed that Turkish delight made with runs 19 and 20 (80.0 g/Kg sugar + 10.0 g/Kg starch + 115.0 g/Kg citric acid) and 75 g/Kg sugar + 11.0 g/Kg starch + 121.7 g/Kg citric acid) achieved the highest score. On the other hand, run 1 (75 g/kg sugar + 12.68 g/kg starch + 117.5 g/kg citric acid) had the lowest score. Application of response optimizer (RSM) yielded the optimal blend of 76.61 g/Kg sugar + 10.72 g/Kg starch + 121.7 g/Kg citric acid which was extremely near to run 19. The high accuracy from the Support Vector Classifier (SVC) was validated by classical machine learning, albeit it varied depending on the trait. Recall scores, in particular, were better for the Variational Quantum Classifier with Nelder-Mead optimization. The study ushers in a new era for the application of quantum machine learning in the food sector.