“…If the inputs are of different scales, the weights linked to some inputs will be updated much faster than others, which is undesirable. Hence, they are usually linearly normalized to be in the range of either [0, 1] or [−1, 1], using [ 97 , 98 , 99 ]: where r i is the particular input data, r min is the smallest input data, and r max is the largest input data, respectively. Various studies have compared ANN algorithms with conventional optimization methods such as the Taguchi method [ 53 , 100 , 101 , 102 ], polynomial regression model [ 67 , 103 , 104 ], and ANOVA [ 46 , 102 , 105 ].…”