<![CDATA[Melon (Cucumis melo L.) is one of the favorite fruits in Indonesia. The relatively short harvest period and high price of melons make melons a superior business commodity. Quality testing of melon fruit is still widely done by relying on destructive testing, which damages the fruit. Therefore, a non-destructive testing approach is needed to predict the parameter values of the physicochemical properties of melon fruit non-destructively with acoustic impulse technology. This study aims to develop a model to predict the physicochemical properties of melon based on parameters of acoustic properties. A total of 120 Honey Globe melons (Cucumis melo var. inodorus) cultivated in the Greenhouse FRC UGM were used as samples. Each fruit was measured non-destructively using knocking tools to generate data on dominant frequency, magnitude, zero-moment power (Mo), and short-term energy (STE). Destructive testing was subsequently conducted to measure moisture content, total soluble solids, and hardness. The destructive and non-destructive test data obtained were processed using an artificial neural network (ANN) to build a prediction model. The training algorithm used was Backpropagation. The results of the ANN training showed the best network structure was 4-4-1. The best learning rates used are 0.1 and 1. Analysis of the reliability of predictions using artificial neural networks carried out based on the calculation of R2 and Mean Squared Error (MSE) values shows that the prediction model consisting of model I, model II, and model III can fulfill the predictions made with high R2 test values, which are sequentially 0.98875; 0.96716; and 0.9215; and MSE values that are relatively small, which are sequentially 0.0016; 0.5296; and 0.2002.]]>
