Comparative analysis of machine learning algorithms within the domain of charitable giving
LE3 .A278 2022
Bachelor of Computer Science
Massive data collection combined with advancements in computing technologies have rapidly progressed the field of predictive analytics. Modelling techniques which previously depended almost entirely on human input are now heavily automated. However, even with this automation, discretion is still often left to the data scientist to determine which techniques best suit a given purpose. Large discrepancies exist between the performance of machine learning algorithms, dependent upon the underlying data and application. One potential application for predictive analytics is in the field of charitable giving. Here, fundraisers must prioritize the allocation of their time and resources based upon the likelihood and significance of prospect donations. Similar methods can also be used to maintain rapports with established donors, by predicting when they might be likely to lapse, or which lapsed donor relationships might be re-established. Thus, a variety of scenarios exist within the field of charitable giving which stand to benefit from predictive analytics. In this thesis, we seek to perform a comparative analysis of an assortment of common machine learning algorithms in the context of charitable giving. These algorithms will be used to predict prospect/donor behaviour across four areas of interest: prospects who are likely to make a donation within six months, donors likely to lapse, previous donors likely to return, and one-time donors likely to make repeat donations.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.