Credit Card Fraud Detection

Inspiration

Every year, thousands of transactions are stolen by cyber criminal and this is creating concern among bank users. Therefore I want to build a model to detect fraud transaction so that it can create an alert to the bank and its users.

What it does

It takes a large imbalanced dataset

How I built it

I use mostly Python and its library to create Classification model.

Challenges I ran into

My dataset is very imbalanced so I have tried SMOTE to oversampling data and made it more balanced. Also, my variables are PCA transformed and due to privacy reason, I was not able to know which variable telling therefore this made it harder for me to control variables.

Accomplishments that I'm proud of

After trying different methods I have found Logistics Regression works well with f_1 score of 73%, accuracy of 97.9% and the confusion matrix with the lowest number of False Negative.

What I learned

This is my first ML project so I have learnt a lot about data preprocessing, creating Classification method and how to maximize the score and confusion matrix.

What's next for Credit Card Fraud Detection

I want to build a website that has different sliders so that users can try out to see how different information will give a different prediction

Built With

• kera
• numpy
• python
• scikit-learn