Abstract: Decision tree is widely used in data mining which is one of the typical classification algorithms. Traditional ID3 tree learning algorithms require training data to reside in memory on a single machine, so they cannot deal with massive datasets. To solve this problem, this paper analyzes the parallel algorithm of ID3 decision tree based on MapReduce model, then proposes a parallel and distributed algorithm for ID3 decision tree learning. The experimental results demonstrate the algorithm can scale well and efficiently process large-scale datasets on commodity computers.

Key words: cloud computing, data mining, decision tree, ID3, MapRedue