Fig 1.
Schematic of multilayer perceptron neural network (MLPNN) architecture.
Fig 2.
Schematic representation of genetic algorithm (GA) as an evolutionary optimization algorithm.
Fig 3.
Flowchart of integrating multilayer perceptron neural network (MLPNN) with genetic algorithm (GA) to optimize MLPNN architecture.
Fig 4.
The characteristics of Nicotiana tabacum cultivars, including Bergerac, Bell, Burly, and Basma, in 2015 growing season.
Mean values are given, standard error are represented by vertical lines. Means followed by the same letter are not significantly different (p ≤ 0.05).
Fig 5.
The characteristics of Nicotiana tabacum cultivars, including Bergerac, Bell, Burly, and Basma, in 2016 growing season.
Average values are given, standard error are represented by vertical lines. Means followed by the same letter are not significantly different (p ≤ 0.05).
Table 1.
Statistics on multiple linear regression (MLR), stepwise regression (SR), principal component regression (PCR), ordinary least squares regression (OLSR), partial least squares regression, and multilayer perceptron-genetic algorithm (MLPNN-GA) for modeling leaf quality responding to blue mold severity, chlorophyll, nitrogen, sugar, nicotine, chloride, and potassium content in four tobacco cultivars “Bergerac, Bell, Burly, and Basma”.
Table 2.
Importance (according to the sensitivity analysis) of the input variables, including blue mold severity, chlorophyll, nitrogen, sugar, nicotine, chloride, and potassium content, for the achievement of the maximum leaf quality in four tobacco cultivars, including Bergerac, Bell, Burly, and Basma, using multilayer perceptron neural network-genetics algorithm models (MLPNN-GA).