大数据下智能手机用户行为特征与选择偏好预测  

Smartphone Users’ Behavior Features and Choice Preference Forecast Under Big Data Background

作　　者：程豪 吕晓玲[2,3] 范超[3] 赵昱 

机构地区：[1]中国科学技术协会创新战略研究院调查统计中心,北京100012 [2]中国人民大学应用统计科学研究中心,北京100872 [3]中国人民大学统计学院,北京100872 [4]Quest Mobile&人大统计移动互联网大数据研究院,北京100015

出　　处：《统计与决策》2018年第2期73-77,共5页Statistics & Decision

基　　金：中央高校基本科研业务费专项资金资助项目(2011030017)

摘　　要：为了避免敏感信息对用户行为特征研究的局限,响应手机智能化脚步,文章提出以APP应用程序为标识的分析思路,借助决策树和其他七个分类模型,刻画用户行为特征,完成选择偏好预测。研究表明,决策树建立手机属性与APP应用程序间的关系,发现不同类型的手机用户行为特征规律。经过对比研究不同手机类型APP选择偏好的分类准确性,选择预测效果最佳模型——随机森林和Bagging,为智能手机由研发阶段投入市场环节提供方法学依据。To avoid the limitations of sensitive information on users＇ behavior features and keep pace with mobile intelli- gence, this paper comes up with a new thought marked by APP application program and employs decision tree and other seven classification models to depict users＇ behavior features and forecast choice preference. Study result shows that decision tree builds a relationship between mobile attributes and APP application and finds the law of user behavior features. Through comparative analysis on the classification accuracy of choiee preference to the smart-phone APP of different types, the paper selects a model with best performance in forecast--Random Forest and Bagging, which provides a methodology foundation from smart-phone de- velopment stage to its marketing links.

关 键 词：智能手机 APP应用程序 用户行为特征 选择偏好预测 

分 类 号：C811[社会学—统计学]