Research Of The C4.5Algorithm Based On CUDA

The21st century is the era of information exp losion, how to use the existinginformation and data effectively and find the law will undoubtedly become an importantissue. At present, many companies or businesses have huge amounts of data stored inthe database, such as customer spending records and customer registration records ofthe food and beverage industry; customer calling and SMS records of network operators;sales records of large supermarkets, etc. finding law from these massive data, and thenguide the decision-makers to make decisions is the task of data mining.Selecting the appropriate data mining algorithms is necessary. There are manytypes of data mining algorithms, including association rules, cluster analysis,classification analysis, etc. Apriori algorithm is the most classical association rulealgorithm, in this paper FP-growth algorithm is discussed based on the apriori algorithm,and gives the parallel computing ideas of FP-growth algorithm; the section of clusteringalgorithm highlights K-MEANS algorithm and gives its parallel processing ideas;decision tree algorithms is more important in classification analysis, this paper focuseson the C4.5algorithm and mainly introduces evaluating tests,handling unknownattribute values, windowing, grouping attribute values, pruning decision trees and otherissues. Its parallel processing ideas is given based on discussing it in detail. With theidea of parallel processing, this paper describes the implementation of the parallel C4.5algorithm with CUDA. CUDA is a platform of general computing of GPU, using theextended C programming language and the compiler of nvcc. Experiments show that theparallel C4.5algorithm with CUDA greatly enhances the efficiency of the C4.5algorithm.This paper confirmed the efficiency of GPU computing, parallel data miningalgorithms with CUDA will effectively enhance the efficiency of data mining.