Table 1.

The research direction of metalearning in big data.

Research directions	Application scenarios	Research content	Learning objects	Advantages	Disadvantages
Classifier	Data prediction	Decision tree Logistic regression Naive Bayes Neural networks	Classification model with metafeatures	High prediction accuracy	Poor performance in schemes with strong indicator dependence
Metric	Few-shot learning	Relation network [30] Neural network [30] Prototypical network [31] Siamese neural network [32]	Metric space	Learning in space is efficient	Not applicable to regression and reinforcement learning
Optimizer	Finding the best strategy	Matching network [33] Graph neural network [34] Gradient descent [36] Curvature information matrix [37]	Optimizer	Metalearner can independently design an optimizer to complete new tasks	High optimization cost