AlexLazar1/R-team-project

Team project, where we applied different algorithms on a data set to predict credit cards frauds. The dataset was already computed through a PCA algorithm so most of the features where unknown to us (probably as a security measure) except the timestamp and the amount of the transaction. We applied and compared multiple algorithms such as: Naive Bayes, SVM, Boosting, Random forest then Decisional tree, Linear model and a neural network and concluded that the neural network obtained the best results. My contribution as a member was on making and tunning random forest and decisional tree algorithms. Another thing to mention is that the data set was highly unbalanced (number of frauds very small compared to too non-frauds) so we also plotted the ROC-AUC curve too check the results of our algorithms. This project was made in R.

R