Transcript Data Mining
Predicting Blood Donor’s Attrition with Data Mining Methods Xing Wan Introduction Blood supply in the U.S. is always inadequate Recruitment and retention of blood donor has become a top priority RESEARCH QUESTIONS What kind of blood donors is more likely to drop out? What are critical factors may lead to donor attrition? Introduction BRIEF LITERATURE REVIEW Current situation of blood supply in the U.S. Prior survey based researches found some important factors (such as donation experience and convenient donation place) may influence donor’s decision to return A recent study on freshmen student attrition used data mining techniques DataSets Donation Records Disaster relief appeals and incentive info. Possible survey data Whole Dataset For Modeling Methodology and Tools CRISP-DM Business Understanding Data Understanding Data Preparation Modeling Evaluation (10-fold cross-validation approach) Deployment Tools: SAS DM SAS EG Base SAS Procedure Data preprocessing Dependent Variable Target_dropout: code it as “1” if a donor did not return after his/her first donation Modeling Popular classification methods Logistic Regression; Decision Tree; ANN; SVM Ensemble techniques Bagging ,Busting and Information fusion Procedure Sensitivity analysis measures the importance of independent variable based on the change in modeling performance that occurs if this variable is not included in the model. The greater the performance decrease, the greater the ratio of importance. Results Identify potential dropout by using data mining model Identify important independent variables or predictors Develop and deploy Detention strategies References 1. Cheng, E., Chany, C., & Chauz, M. (2010). Data Analysis for Healthcare: A Case Study in Blood Donation Center Analysis. Americas Conference on Information Systems. 2. Delen, D. (2010). A comparative analysis of machine learning techniques for student retention management. Decision Support Systems, 49, 498-506. 3. Masser, B., White, K., Hyde, M., & Terry, D. (2008). The Psychology of Blood Donation: Current Research and Future Directions. Transfusion Medicine Reviews, 22(3), 215-233. 4. Saltelli, A. ( 2002). Making best use of model evaluations to compute sensitivity indices, . Computer Physics Communications 145, 280-297. References 5. Schlumpf, K., Glynn, S., Schreiber, G., & Wright, D. (2008). Factors influencing donor return. TRANSFUSION, 48, 264-272. 6. Schreiber, G., Sanchez, A., Glynn, S., & Wright, D. (2003). Increasing blood availability by changing donation patterns. TRANSFUSION, 43, 590-597. 7. SPSS. SPSS PASW Modeler (formerly Clementine) User Manual. A Comprehensive Data Mining Toolkit, 2010. 8. Yu, P., Chung, K., Lin, C., Chan, J., & Lee, C. (2007). Predicting potential drop-out and future commitment for first-time donors based on first 1.5-year donation patterns: the case in Hong Kong Chinese donors. Vox Sanguinis, 93, 57–63.