A carregar...
Publicação
Understanding and Predicting Micro-Donations in Washington, D.C. Political Campaigns: An Explainable Machine Learning Approach
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.64 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Political campaign financing has grown increasingly reliant on small-dollar micro-donations,
which broaden the donor base and democratize fundraising. This thesis investigates how
explainable machine learning can predict micro-donation behavior and identify its key drivers,
focusing on political contributions in Washington, D.C. A comprehensive dataset of campaign
finance records is compiled and integrated with socio-demographic and geographic open data
to capture contextual factors and uncover spatial patterns in giving. Following the CRISP-ML
framework, we develop a gradient-boosted tree model (XGBoost) to predict the amount of
micro-donations contributions, using Optuna for rigorous hyperparameter tuning. To ensure
transparency in this predictive approach, we apply SHAP (Shapley Additive Explanations) to
interpret the models’ outputs, allowing us to assess the influence of individual donor
characteristics and donation attributes on the predictions. The results demonstrate high
predictive accuracy, with a Mean Absolute Error (MAE) of approximately $1.56 and an R² of
0.8. The results show that contributor identity is the primary driver of micro-donation
behavior, followed by committee and candidate names, donation patterns, and the
demographic and socioeconomic contexts. These insights highlight the strong role of
individual donor behavior while also capturing the influence of organizational ties and
neighborhood characteristics. Notably, the inclusion of demographic and geographic data
provides valuable insights into community-level trends and patterns, enhancing our
understanding of how local context shapes political giving. By combining strong predictive
performance with interpretable insights, this study offers a novel, data-driven perspective on
the mechanics of small-scale political contributions. The findings enrich current debates on
campaign finance by highlighting the significance of micro-donations and providing evidencebased recommendations for campaign strategies and policy reforms that aim to enhance
transparency, promote equity, and encourage broader civic participation in the political
process.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Business Intelligence
Palavras-chave
Political Contributions Campaign Finance Micro-Donations Donor Behavior Washington, D.C. Open Data Machine Learning Explainable AI (XAI) SHapley Additive exPlanations (SHAP) XGBoost Optuna SDG 10 - Reduced inequalities SDG 11 - Sustainable cities and communities SDG 16 - Peace, justice and strong institutions SDG 17 - Partnerships for the goals