In this digital world, most of the transactions are occurring via credit cards while using cash will become rare in future days. At the same time, the information system is putting more efforts to make the credit card system more reliable and trustworthy. AI has made this task as easy as it can be and Python data science packages have made it easier for data analysts. Using the methods followed by the data scientists, I tried to detect the fraud by exploring the dataset. Fraud detection in credit cards is necessary for the future as the hackers always look for vulnerability so we must update our ideas and AI machine learning has created the opportunity to do so.

What it does

It just explores the dataset, cleans it, and by running the classifiers detects the rows as fraud or not a fraud.

How I built it

I used Jupyter Notebook and Python data science packages to build it. Numpy, pandas, scikit learn, matplotlib, seaborn, etc. data analysis and data visualization packages have been used throughout the project.

Challenges I ran into

Data cleaning is an important task to do in this kind of project and it was tough to detect the outliers for such a big dataset. The cleaning of null values, irrelevant, and outliers were challenging.

Accomplishments that I'm proud of

I had got really a little time to work on this big dataset yet have completed successfully what I have intended to do. This is a real-world project which can contribute to better human life. I'm really proud of what I've made despite knowing that I could do more better and investing a lot of time in building an ML project is really praiseworthy. I had to seek help from the developers to fix bugs and errors while data explorations which gave me a lot of insights regarding the techniques to be used in pandas and NumPy.

What I learned

I learned how to work with sklearn packages and which classifiers work better in which condition and discovered the latency of the pandas package.

What's next for AI implementation in Credit Card Scam Detection

It's just the starting; it can be developed for making future sophisticated machines and AI ML models. The more open source datasets must be available for the better training of the models. The scarcity of open-source datasets is also challenging for the development of such models.

Built With

Share this project: