This simple machine-learning project finds outliers in transactional finance data.
It does so by utilizing a decision tree classifier combined with the boosting algorithm AdaBoost.
This repository contains the steps of developing this method and exploring other options.
The notebook training.ipynb contains the analyis of the data and an exploration of different methods to solve the problem like Neural Networks, Random Forests, Logistic Regression and Decision Trees. It also tests the finally selected method and determines the accuracy on the testing and training data.
Accuracy Score on testing data: 97%
Sensitivity: 100% Specificity: 99%