Predicting Eggplant Individual Fruit Weight Using an Artificial Neural Network

Abstract

ABSTRACTEstimation of relationships between inconstant factors can be helpful to calculate amounts of variation of a particular character with respect to others. In Eggplant (Solanum melongena L) this information could be used to improve fruit yield. Effects of agronomic and phenologic factors were studied by applying an artificial neural network (ANN) as a displaying instrument to determine how plant length; individual fruit weight, length, and width; number of fruit per plant; ratio of fruit length to fruit width; total yield; number of days to flowering; number of days to first harvest; canopy temperature; chlorophyll; and relative water content affected individual fruit weight of eggplant. There was a high accuracy obtained for the 7-4-1 ANN model based on these parameters (R2 = 93%; mean prediction error [MPE] = 2.01; mean square deviation [MSD] = 2.35). A sensitivity analysis was performed and the ratio of fruit length to fruit width, number of days to first harvest, and number of days to flowering had the greatest impact on individual fruit weight. The highest standard deviation was for total yield and individual fruit weight, respectively (308.8 and 67.5), and correlation coefficients were high between fruit weight and number of days to flowering (0.99**) and individual fruit weight and total yield (0.88**). Sensitivity analysis indicated that the ratio of fruit length to fruit width and fruit length have high and lesser effects on final individual fruit weight. Total yield is the main factor for producing change in individual fruit weight.