Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Date of Graduation
Spring 5-7-2010
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Department of Integrated Science and Technology
Abstract
The objective of this thesis is to compare the predictive performance of multiple regression, logistic regression, neural networks, and support vector machines (SVM) in identifying donors and non-donors, and predicting the amount of donations for a specific solicitation campaign using a data set from the Direct Marketing Education Foundation (DMEF). Non-Profit fundraising has benefited from the use of data-mining models to identify new donors, to re-solicit existing donors, and to increase the amount of donations from a solicitation campaign. Multiple regression, logistic regression, and neural networks have been widely used for non-profit fundraising. In this thesis the support vector machines are used to identify new and repeated donors, and to predict the amount of donations. The SVM models have been used mostly in machine learning and other non-business applications. The SVM models have several advantages over multiple regression, logistic regression, and neural networks. The SVM models are free from the curse of the dimensionality as in both multiple regression and logistic regression, and are also free from local minimums during training as in neural networks. The SVM models are optimization models with easy interpretations and the potential capability of handling hundreds of thousands of variables. The thesis will show how the four methods are used and will discuss the pros and cons of using each of the methods for non-profit fund raising.
Recommended Citation
Chen, Qin, "Predictive modeling for non-profit fundraising" (2010). Masters Theses, 2010-2019. 415.
https://commons.lib.jmu.edu/master201019/415