Enhancement of nutritional value of fried fish using an artificial intelligence approach

Abstract

Frying affects the nutritional quality of fish detrimentally. In this study, using Catla catla and mustard oil, experiments were carried out in varying temperatures (140-240 °C), times (5-20 min), and oil amounts (25-100 ml/kg of fish) which established drastic reduction of 44.97% and 99.40% for polyunsaturated fatty acid (PUFA)/saturated fatty acids (SFA) and index of atherogenicity (IA) profile, respectively. Artificial neural network (ANN) was implemented successfully to provide an association between the independent inputs with dependent outputs (values of R² were 0.99 and 0.98; RMSE were 0.038 and 0.046; and performance were 0.038 and 0.067 for PUFA/SFA and IA, respectively) by exhaustive search of various algorithms and activation functions available in literature. ANN model-based meta-heuristic, stochastic optimization formalisms, genetic algorithm (GA) and particle swarm optimization (PSO), were applied to optimize the combination of cooking parameters for improving the nutritional quality of food which improved the nutritional value by maximizing the PUFA/SFA profile up to 63.05% and minimizing the IA profile to 99.64%. Multi-objective genetic algorithm (MOGA) was also employed to tune the inputs by maintaining a balance between the contrasting outputs and enhance the overall food value simultaneously with multi-objective (beneficial for health, economic, and environment-friendly) proposal. MOGA was able to improve the PUFA/SFA profile up to 44.76% and reduce the IA profile to 92.94% concurrently with the reduction of wastage of culinary media and energy consumption, following the optimized cooking condition (118.92 °C, 6.06 min, 40 ml oil/kg of fish).

Keywords: Artificial neural network; Fish; Frying; Genetic algorithm; Multi-objective genetic algorithm; Nutritional value; Particle swarm optimization.