MapReduce implementation for minimum reduct using parallel genetic algorithm

Rough set theory (RST) proved to be an effective approach in data mining which can be used successfully for feature/attribute selection and rule induction. Unfortunately, the search space created by RST can be huge and it is important to reduce the search time for the shortest reduct. Genetic algorithm (GA) is one of the metaheuristic algorithms that have been used to tackle this NP-hard optimization problem. However, the effectiveness of the genetic algorithm depends on its implementation. In this work, we introduce a MapReduce approach of a parallel generic algorithm to find the minimum reduct. We evaluated the proposed approach on a number of cybersecurity datasets with varying characteristics. The results showed that the MapReduce approach was more efficient than the sequential approach especially when we go for high dimensions.