Abstract A novel Partial Effect Regression and Convergence-efficient Deep Belief Neural Network (PER-CDBNN) method is proposed based diet plan recommendation for big data analytics in child nutrition with several layers. In input layer, diet plan samples from corresponding big dataset are provided as input. Then, first hidden layer of deep belief neural network, Peirce’s Criterion Probability-based pre-processing is applied to ensure robust cleaning and normality of big data. Next second hidden layer, a feature selection model called, Symbolic Regression and Partial Effect is applied by revealing inputs possessing most significant unique influence on outcome. Finally, to balance learning speed and prevent slow convergence, Convergence-efficient ReLU activation-based Classifier is employed to multiclass predictions for accurate diet plan recommendation. These uncovering constitute a paradigm shift toward proactive nutritional surveillance, enabling diet recommendation through Deep Belief Neural Network methods. Implementation of this could significantly boost early recommendation capabilities, clear way for timely nutritional interventions and contribute considerably to bringing about nutrition targets globally while reducing childhood mortality rates. Keywords: Deep Learning, Peirce’s Criterion Probability, Symbolic Regression, Partial Effect, Convergence-efficient ReLU
Vanitha et al. (Sun,) studied this question.