Question

Credit Card Fraud: From a dataset of the form [Account, Company, Date[DD/MM/YYYY], Fraudulent charge [boolean]] find the merchant that was originally compromised in the dataset. Look for a large number of accounts interacting with the company a few days before fraudulent charges are seen in the accounts. Using python, what approach would you use? Name any packages/ ML libraries or other that would be utilized.

Answer 1

According to the model proposed in the blog named "In depth skewed data classif. (93% recall acc now)" written by joparga3In.

By using one of the famous technique called cross-validation technique, we can choose  penalty parameter of logistical regression such that the recall score is optimized. By finely tuning the penalty parameter gives boost to obtain high recall score.

So using well trained logistic regression can be useful to classify the fraud activities.

We can run the algorithms using different thresholds.

Following packages or libraries are sufficient

1) pandas matplotlib

2) sklearn

3) scipy