区块链框架下基于优化决策树模型的大数据分类算法研究  

Research on Big Data Classification Algorithm Based on Optimized Decision Tree Model Under Block Chain Framework

作　　者：杨薇薇[1] 曾凌静[1] YANG Wei-wei;ZENG Ling-jing(School of information and intelligent transportation,Fujian Chuanzheng Communications College,Fuzhou 350007,Fujian Province)

机构地区：[1]福建船政交通职业学院信息与智慧交通学院,福建福州350007

出　　处：《沈阳工程学院学报（自然科学版）》2021年第4期57-62,共6页Journal of Shenyang Institute of Engineering:Natural Science

基　　金：2018福建省中青年教师教育科研项目(JZ180372)。

摘　　要：现有大数据分类算法存在并行计算能力差、分类准确率低等不足,为此在区块链框架体系下提出一种基于优化决策树模型的大数据分类算法研究。利用区块链在数据加密方面的优势,构建分布式决策树模型,以信息熵和信息增益率为基础对目标数据集做预测分类,并对经典决策树算法进行优化,预留出用于缓冲的空节点,避免分类终止情况的发生。在每一次节点分裂后,用全部样本的平均梯度修改下一个叶子节点的权重,提升整个算法的迭代寻优能力和分类性能。仿真结果显示,提出的分类算法具有更强的数据吞吐能力和并行计算能力,针对10种不同大数据集的平均分类准确率达到了97.75%。The existing big data classification algorithms had poor parallel computing ability and low classification accuracy.Therefore,a big data classification algorithm based on optimized decision tree model was proposed under the block chain framework.Based on the advantages of block chain in data encryption,a distributed decision tree model was constructed.Based on the information entropy and information gain rate,the target data set was predicted and classified.The classic decision tree algorithm was optimized to reserve the empty nodes for buffering to avoid the occurrence of classification termination.After each node splitting,the average gradient of all samples was used to modify the weight of the next leaf node to improve the iterative optimization ability and classification performance of the whole algorithm.Simulation results show that the proposed classification algorithm has stronger data throughput and parallel computing capability,and the average classification accuracy reaches 97.75%for 10 different large data sets.

关 键 词：区块链 优化决策树模型 大数据分类 并行计算 平均梯度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]