Effects of Feature Types on Donor Journey

A donor journey is the series of actions taken by a fundraising institution and one of its donors that leads to a donation. This journey can take place over a few seconds or a few years and can involve one action (donation) or hundreds of actions, ending with a donation. Ultimately, fundraising institutions are most interested in the next best action to take that will lead to a significant donation. In the past, research has been done on this topic to discover which combination of features of actions and constituents can be used to best predict which series of actions will lead to a donation. This research involves the use of temporally capable deep learning algorithms (e.g., Recurrent Neural Networks) and other forms of deep learning (e.g., Convolutional Neural Networks). We extend this work by combining constituent features with time-based features and show that the combination leads to machine learned models with lower MAEs than previously created models, ultimately resulting in a model with an MAE of $22 for a wildlife charity.