In this project, we make use of the binary classifier we trained on the transactions dataset to detect fraudulent transactions.
In the leader_board_predict_fn function contains the code that loads our model parameters and returns the likelyhood of fraud for each transaction (i.e. row) in the values dataframe.
The higher the decision function value, the more likely that the transaction is fraud.
We used three different machine learning models and trained them with the given transactions data, and yielded the score,
to find the model that showed best performance.
*details are in report.pdf