Abstrait

Wavelet neural network based on genetic algorithm for modeling enzymatic esterification of betulinic acid using phthalic anhydride as acylating agent

Mansour Ghaffari-Moghaddam, Mansoureh Rakhshanipour, Mostafa Khajeh, Faujan Bin H.Ahmad

In this study, a wavelet neural network (WNN) constructed of general neural network employing the wavelet function as the activation function was used in the enzymatic synthesis of betulinic acid ester using phthalic anhydride as acylating agent. The genetic algorithm (GA) was selected to optimize the weights of neural network. The input parameters of the model were reaction time, reaction temperature, amount of enzyme and substrate molar ratio while the percentage isolated yield of ester was the output. After evaluation of various WNN configurations, a topology with 4-15-1 arrangement gave the best performances. The root mean square error (RMSE) and coefficient of determination (R2) between the actual and predicted yields were determined as 1.8366 and 0.9758 for training set, 0.7915 and 0.9976 for testing set and 4.1991 and 0.8339 for validation set, respectively. The constructed WNN-GA model showed relatively higher importance of time and amount of enzyme than temperature andmolar ratio in the enzymatic reaction.All these results showed that theWNN-GAhas a great potential ability in predicting the isolated yields of the enzymatic reaction.