基于分布式运算的决策树算法的研究与实现

Research and Implementation of Decision Tree Algorithm Based on Distributed Computing

摒弃决策树算法中一般常用的递归运算,利用网络中的任意台计算机,采用C/S结构模式,由服务器主要存储决策树的节点信息,客户机并行计算出决策树的分割属性及其分支,以解决计算过程中的内在崩溃现象。实践表明,C/S结构模式可建立于局域网、广域网,且可充分利用云计算平台上的高性能虚拟网络,具有结构简单、灵活,适应面广等特点,在相同样本数据集上,测试正确率较高,算法切实可行。

The paper abandons the recursive algorithm commonly used in decision tree algorithm. Using the C / S structure model in the network, the server mainly stores the node information of the decision tree, and the client calculates the segmentation attribute of the decision tree and its branches in parallel to solve the inherent collapse phenomenon in the calculation process. The practice shows that the C / S structure model can be built on the local area network and the wide area network, and can make full use of the high performance virtual network on the cloud computing platform. It has the characteristics of simple structure, flexibility and wide adaptability. On the same sample data set, the test accuracy is high and the algorithm is feasible.