August 2019
Intermediate to advanced
342 pages
9h 35m
English
Among the oversampling methods, we have the Synthetic Minority Over-sampling Technique (SMOTE); this allows for the generation of synthetic samples by interpolating the values that are present within the class subjected to oversampling.
In practice, synthetic samples are generated based on the clusters that are identified around the observations present in the class, therefore calculating the k-Nearest Neighbors (k-NNs).
Based on the number of synthetic samples that are needed to rebalance the class, a number of k-NN clusters are randomly chosen, around which synthetic examples are generated by interpolating the values that fall within the selected clusters.
Read now
Unlock full access