In this project we sought to optimize the outreach strategy for a home repair tax credit program by employing data-driven methods. The primary objectives were to identify eligible homeowners, predict their likelihood of participating in the program, and evaluate the cost-effectiveness of the outreach efforts. By utilizing historical data from previous outreach campaigns, the project aimed to develop a classification model that could accurately sort homeowners into two categories: those likely to participate and those unlikely to participate. This predictive model would serve as the foundation for a more targeted and efficient outreach strategy, ultimately increasing the number of participants while minimizing marketing expenses. To achieve these goals, the project involved a comprehensive analysis of key variables, such as the inflation rate, the number of previous contacts, and the unemployment rate on the day of contact. The data exploration phase revealed notable correlations between these variables and program participation. The project then focused on building and comparing two logistic regression models: a "Kitchen Sink" model incorporating all available features and an "Engineered" model with carefully selected and transformed variables. Through cross-validation techniques and cost-benefit analysis, the project evaluated the performance of each model in terms of accuracy, sensitivity, specificity, and financial implications. The findings highlighted the importance of balancing the dataset and refining the model to improve its ability to identify potential participants effectively. The project concluded with recommendations for enhancing the outreach strategy, including improved sampling techniques, model refinement, and targeted marketing efforts, to maximize the program's impact and cost-effectiveness.
