基于随机森林的Titanic乘客生存预测  

Titanic Passenger Survival Prediction Based on Random Forest

作　　者：苗水清 吴梦蝶 吴晓晖 肖晗昊 田松洁 MIAO Shui-qing;WU Meng-die;WU Xiao-hui;XIAO Han-hao;TIAN Song-jie(Xian Innovation College of Yan'an University,Xi'an 710100,China)

机构地区：[1]延安大学西安创新学院,陕西西安710100

出　　处：《山东工业技术》2020年第4期68-72,共5页Journal of Shandong Industrial Technology

基　　金：高校图书馆选书决策方案研究,2019XJKY-1,校级。

摘　　要：本文基于Titanic数据生存预测问题展开了分析和研究,首先对Titanic数据进行预处理,选择对结果影响最为直观的特征量,然后采用随机森林算法在多种超参数设置下进行实验,结果表明:采用随机森林在通过调整超参数后能够得到较好的分类效果,正确率能达到93%以上,同时与其他算法进行了对比,验证了随机森林算法是比较理想的预测方法。Based on the binary classification problem of Titanic data, this paper has carried out analysis and research. The specific ideas include: first preprocess the titanic data, select the feature that has the most intuitive effect on the result, and then use the random forest algorithm to experiment with multiple hyperparameter settings, And the results show that the random forest can get a better classification effect after adjusting the hyperparameters, and the accuracy rate can reach more than 93%. Compared with other classification algorithms, this method has the advantages of high recognition rate, short time consumption, and is an ideal prediction method.

关 键 词：随机森林 预测 数据集 特征量 

分 类 号：TP3-05[自动化与计算机技术—计算机科学与技术]