Today, we are living in a data-exploding era, in which the volume of data is expanding in an unbelievable fast way and the speed is faster than any period in the history. Using machine learning algorithms for processing massive data has become a hot research area now and lots of computer scientists and developers use them to extract hidden information from massive data. However, as the volume of data has increased too much for recent years and the trend is still increasing, just by using a standalone machine to deal with these massive data is becoming unrealistic as the volume of data and the computing complexity for processing massive data has exceeded the capacity of a single machine. In order to solve this problem, in this paper, we combined Extreme Learning Machine(ELM), which is a machine learning algorithm that has the ability of extreme fast training, and mapreduce parallel framework to proposed a mapreduce-based ELM called MR_ELM. And according to some experiments by using KDDCUP99 dataset, we have proven that MR_ELM can process massive data in a parallel way without losing accuracy performance compared with local ELM.
