Prediction of apple drying level by machine learning and electronic nose

Abstract In this study, changes in the odour of apple slices throughout the drying process were monitored and documented using electronic nose sensors, while moisture levels were measured and recorded using an electronic hygrometer. The initial and final moisture contents of apple slices were determined as 86.81 ± 0.05% and 4.92 ± 0.01%, respectively. During drying, apple slices were monitored with six odour sensors powered by an Arduino microprocessor. As the moisture value of the apple slices decreased during the drying process, the electronic nose data also decreased. In addition, the image changes in the apple slices according to the drying level were taught to the Teachable Machine. In this study, it was observed that the machine was able to detect the drying level of apples with 85%–100% accuracy. When 900 apple samples were used in the test phase, machine learning was able to predict the drying level of apples with 100% accuracy based on the confusion matrix data. PCA analysis highlighted different patterns in the sensor data during drying; PLS analysis showed that the sensor data can accurately predict the colour (L*, a*, b*), diameter, and thickness of apple slices with high correlation coefficients (r CV and r Pre > 0.9).