Approx-SMOTE: Fast SMOTE for Big Data on Apache Spark

One of the main goals Big Data research, is to find new data mining methods that are able process large amounts in acceptable times. In classification, as traditional class imbalance a common problem must be addressed, case also looking for solution can applied an execution time. this paper we present Approx-SMOTE, parallel implementation SMOTE algorithm Apache Spark framework. The key difference with original SMOTE, besides parallelism, it uses approximated version k-Nearest Neighbor which makes highly scalable. Although already exists (SMOTE-BD), exact Nearest search, does not make entirely Approx-SMOTE on other hand achieve up 30 times faster run without sacrificing improved classification performance offered by SMOTE.