Distributed Lazy Association Classification Algorithm Based on Spark

The lazy association classification algorithms are inefficient when classifying multiple unclassified samples at the same time. The existing lazy association classification algorithms are sequential which can’t deal with the big data problems. To solve these problems, we propose a distributed lazy association classification algorithm based on Spark, named as SDLAC. Firstly, it clusters the unclassified samples by K-Means algorithm. Secondly, it executes distributed projections according to clustered results, and mines classification association rules by a distributed mining algorithm based on spark. Then it constructs classifier to classify unclassified samples. The experiments are conducted on the 5 UCI datasets and a big dataset from the first national college competition on cloud computing(China). The results show that SDLAC algorithm is more accurate than the CBA algorithm. Besides, its efficiency is far more than the typical distributed lazy association classification algorithm. In other words, the SDLAC algorithm can adapt big data environment.