Predicting the downfall of non-profit organizations using machine learning

Abstract

Machine learning can be applied to finances of non-profit organizations taken from IRS Tax Forms 990ez to determine if an organization will be dissolved. This is useful to determine if a cause is viable. Data stored on an online database is extracted, formatted, parsed and segregated using Python. The code selects the attributes used to predict the organization’s downfall. Finances were compared and attributes that were critical were identified. Three supervised predictive algorithms, Decision Tree, K Nearest Neighbors and Naïve Bayes, were used. Results from the algorithm's predictions for organizations that were dissolved and non-dissolved are presented in this paper and discussed. This study also determined the average duration of non-profit organizations based on the current financials. Key Terms ⎯ Algorithms, Analytics, Big Data, Prediction, Machine Learning.