Fraud detection (classification task) in a context of inbalanced data using Credit card dataset -Unbalanced dataset-.
The deployed web app is live at (Link)
This web application predicts the class of a bank transaction as fraudulent or legitimate, using 4 algorithms (LogisticRegression, DecisionTreeClassifier, RandomForestClassifier, XGBClassifier), and different resampling methods to deal with the imbalance problem. A cost-sensitive method is used to improve the algorithms in order to reduce the number of false classifications (false positives, false negatives).
The web application was built in Python using the following libraries:
- streamlit
- matplotlib
- numpy
- scikit-learn
- pickleshare
- matplotlib
- imbalanced-learn
This research was funded by Grassfields, a consulting, training, and R&D company, as part of an end-of-studies internship. Grassfields played a crucial role in supporting this study, allowing for a deeper exploration of the subject matter. Through their dedication and expertise, this research yielded significant and enriching results. (Grassfields)
Mr Idriss Tchapda Djamen acted both as director of Grassfields and as supervisor of this work.
