Analyzing Taxation Data Using Machine Learning Techniques

Abstract

This aims to contribute to the extant literature in this field by examining nonprofit organizations’ fraud reporting compliance using classification algorithms on the basis of financial indicators. SVM (Support Vector Machine), MLP (Multilayer Perceptron), KNN (K Nearest Neighbor) and Naïve Base to classify the asset diversions by NPOs. The experiments performed on the basis of four training and test split sets that are Split by 90%, 80%, 70% and 60%. Results on the basis of classification process applied on each algorithm are compared and based on criteria Accuracy and F Score the best algorithm is being chosen. NB and SVM have the same criteria F Score and Accuracy and these are the best results among all percentage split experiments.