<![CDATA[Artificial neural networks (ANN) was used to predict the quality parameters of tomato, i.e. Brix, citric acid, total carotene, and vitamin C. ANN was developed from Red Green Blue (RGB) image data of tomatoes measured using a developed computer vision system (CVS). Qualitative analysis of tomato compositions were obtained from laboratory experiments. ANN model was based on a feedforward backpropagation network with different training functions, namely gradient descent (traingd), gradient descent with the resilient backpropagation (trainrp), Broyden, Fletcher, Goldfrab and Shanno (BFGS) quasi-Newton (trainbfg), as well as Levenberg Marquardt (trainlm). The network structure using logsig and linear (purelin) activation function at the hidden and output layer, respectively, and using the trainlm as a training function resulted in the best performance. Correlation coefficient (r) of training and validation process were 0.97 - 0.99 and 0.92 - 0.99, whereas the MAE values ranged from 0.01 to 0.23 and 0.03 to 0.59, respectively.ABSTRAKJaringan syaraf tiruan (JST) digunakan untuk memprediksi parameter kualitas tomat, yaitu Brix, asam sitrat, karoten total, dan vitamin C. JST dikembangkan dari data Red Green Blue (RGB) citra tomat yang diukur menggunakan computer vision system. Data kualitas tomat diperoleh dari analisis di laboratorium. Struktur model JST didasarkan pada jaringan feedforward backpropagation dengan berbagai fungsi pelatihan, yaitu gradient descent (traingd), gradient descent dengan resilient backpropagation (trainrp), Broyden, Fletcher, Goldfrab dan Shanno (BFGS) quasi-Newton (trainbfg), serta Levenberg Marquardt (trainlm). Fungsi pelatihan yang terbaik adalah menggunakan trainlm, serta pada struktur jaringan digunakan fungsi aktivasi logsig pada lapisan tersembunyi dan linier (purelin) pada lapisan keluaran. dengan 1000 epoch. Nilai koefisien korelasi (r) pada tahap pelatihan dan validasi secara berturut-turut adalah 0.97 - 0.99 dan 0.92 - 0.99; sedangkan nilai MAE berkisar antara 0.01-0.23 dan 0.03-0.59.]]>
