摘要
针对传统决策树SPRINT(Scalable Parallelizable Induction of Decision Trees)算法不能处理海量地学数据挖掘的问题,设计实现了基于G4ICCS(Geology Geography Geochemistry Geophysics Information Cloud ComputingSystem)的决策树并行分类算法PSPRINT。该算法使用哈希表存储连续属性分割点两侧的数据记录,为并行节点的分割提供依据,在MapReduce架构下解决了海量地学数据挖掘问题。实验结果表明,在模拟的云计算环境下,决策树并行算法可以处理海量地学数据分类问题,并获得较好的稳定性和较高的处理速度。
For the traditional decision tree SPRINT (Scalable Parallelizable Induction of Decision Trees ) algorithm cannot solve the problem of mass geoscience data mining, the paper designed and realized PSPRINT algorithm. It is a decision tree parallel classification algorithm based on G4 ICES (Geology Geography Geochemistry Geophysics Information Cloud Computing System). The algorithm uses hash table to save data record on both sides of continuous attributes pointof division, providing basis for the division of parallel node, and solved mass geoscience data mining problem. The experimental results show that the decision tree parallel algorithm can deal with the classification problem of mass geoscience data under the simulated environment of cloud computing. And the algorithm has better stability and processing speed.
出处
《吉林大学学报(信息科学版)》
CAS
2013年第3期324-327,共4页
Journal of Jilin University(Information Science Edition)
基金
吉林省"十二五"矿产资源规划预测基金资助项目(3R212H104422)
关键词
地学G4ICCS系统
数据挖掘
决策树算法
并行
geology geography geochemistry geophysics information cloud computing system(G4ICCS)
data mining
decision tree algorithm
parallel