Predictive models for charitable giving using machine learning techniques.

Area of Science:

• Economics
• Data Science
• Sociology

Background:

• Private giving constitutes over 75% of US charitable donations, significantly funding the nonprofit sector (over 10% of GDP).
• Despite available data, factors influencing private donations and predictive models remain underdeveloped.
• Understanding donation drivers is essential for nonprofit sustainability and economic planning.

Purpose of the Study:

• To develop accurate predictive models for future charitable giving.
• To identify key socioeconomic and demographic factors influencing private donation behavior.
• To compare the efficacy of different statistical and machine learning models in forecasting charitable contributions.

Main Methods:

• Utilized Stepwise Regression to identify significant variables from a dataset including unemployment rate, income, poverty, population, demographics, education, and vehicle ownership.
• Developed predictive models using Multiple Linear Regression (MLR), Artificial Neural Networks (ANN), and Support Vector Regression (SVR).
• Validated models using 10-fold cross-validation and evaluated performance with nine distinct metrics.

Main Results:

• Population, education level, and previous year's charitable giving emerged as the most significant independent predictors of future donations.
• All developed models (MLR, ANN, SVR) demonstrated good accuracy in predicting regional charitable giving.
• Artificial Neural Networks (ANN) exhibited superior predictive performance compared to Support Vector Regression (SVR) and Multiple Linear Regression (MLR) on test data.

Conclusions:

• Population, education, and past donation history are critical factors for forecasting charitable giving.
• Machine learning models, particularly ANN, offer a robust approach to predicting future donation amounts.
• Accurate prediction of private giving can aid nonprofit organizations in financial planning and resource allocation.