《r语言学习笔记和2014r会议材料china2014bjlinhui.docx》由会员分享，可在线阅读，更多相关《r语言学习笔记和2014r会议材料china2014bjlinhui.docx（6页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、Analytics to Support Business Decision inTraditional Industry25 May 20147th R conference (Beijing, China)毆 PIONEERTraditional Industry v.s. E-commerce Industrya NOT big dataa Different time seriesa Well defined concentrationa Relationship: Rep Sales play a role Without/little online purchasing data
2、a Machine learning and statistical model a Market research based on questionnaire a Time of the year matters (for agriculture)How Analytics Can Helpa Business Opportunities Cross-selling Evaluate marketing campaign Target specific population (segmentation) R.O.I of different marketing program and se
3、rvices Understand customer attrition/conversion/retention Predict possible customer reaction before decision implementationNot Academica Dirty dataa Useful v.s. Accuratea Issues of observational study Hard to define control group Missing value Correlation? Causality?“Implementation”a Gap between “Im
4、plication” andScoring System for Customer Retention Objective: score customers by the likelihood of retention Response: 0/1 Predictors: categorical, continuous (different scales, all possible historical transactional records) Data Size: 10000 x 200Cross-validdt kn (leave one out, k fold)Group LassoA
5、rea under ROC curveSignificant basedTrain/testOther variableselectionmethods (p-values?)LoglikelihcxxlBootstrapExpertopinionMaximumcorrelation50,000 random t variates, 200 d.f.set.seed(115)-29_dEa)sp_dEe(DNormal distribution quantilesNormal distribution quantilessetseed(116)Anderson-Darling normality testA= 1.1849, p-value = 0.004318Anderson-Darling normality testA= 0.1731, p-value = 0.9281