SMOTE stands for Synthetic Minority Oversampling Technique.
When you train a machine learning model and the input data for classification is very imbalanced (one class 50.000 and the other 1.000) then the outcome of some of the models might be impacted by this. Hence it is recommended to pre-process the data in order to have more balanced data. The reason is that accuracy is often measured with the predictive accuracy, which in such case is impacted by this imbalance.
One common technique used is SMOTE.
In smote synthetic samples are added to the input file in order to address the above issue.
The algorithm is based on N nearest neighbours
SMOTE is part of the pre-processing of the input data.
