Predicting the Chemical Attributes of Fresh Citrus Fruits Using Artificial Neural Network and Linear Regression Models

Different chemical attributes, measured via total soluble solids (TSS), acidity, vitamin C (VitC), sugars (Tsugar), and reducing (Rsugar), were determined for three groups of citrus fruits (i.e., orange, mandarin, acid); each group contains two cultivars. Artificial neural network (ANN) multiple linear regression (MLR) models developed TSS, VitC, Tsugar, Rsugar from fresh by applying different independent variables, namely the dimensions (length (FL) diameter (FD)), fruit weight (FW), yield/tree, soil electrical conductivity (EC). The results ANN application showed that a feed-forward back-propagation type with four input neurons (Yield/tree, FW, FL, FD) eight in one hidden layer provided successful modeling efficiencies Rsugar. effect EC variable was not significant. hyperbolic tangent both output model chosen as activation function. Based on statistical criteria, this study performed better than MLR predicting attributes fruits. root mean square error ranged 0.064 to 0.453 0.068 0.634, respectively, model, 0.568 4.768 0.550 4.830, using training testing datasets. In addition, relative errors obtained through approach high predictability feasibility. attribute modeling, FD FL variables exhibited contribution ratios, resulting reliable predictive model. generally good level accuracy when estimating fruit.