This is a machine learning project of being able to predict whether a transaction is a fraud or not. The Dataset is taken from a closed Zindi competition.
Fraud detection is an important application of machine learning in the financial services sector. This solution will help Xente provide improved and safer service to its customers.
This competition is sponsored by Xente, Innovation Village, and insight2impact.
Xente is an e-payments, e-commerce, and financial services company in Uganda offering various products and services that can be paid for using Mobile Money (Airtel Money, MTN Mobile Money), Bank Card (Visa Card, Master Card), Xente wallet and on credit (Pay Later). Some of the products consumers can buy include airtime, data bundles, pay water and electricity bills, TV subscription services, buy event tickets, movie tickets, bus tickets, and more.
Understanding fraudulent transactions!
- Do fraudulent transactions have a common pattern?
- How do they differ from "normal" transactions?
- What features are important to recognize fraudulent transactions?
Error metric
-
The error metric for this competition is the F1 score, which ranges from 0 (total failure) to 1 (perfect score). Hence, the closer your score is to 1, the better your model.
-
F1 Score: A performance score that combines both precision and recall. It is a harmonic mean of these two variables. Formula is given as: 2PrecisionRecall/(Precision + Recall)
Tech-Stack
- Python, Pandas, Numpy, Matplotlib, Seaborn, Scikit-Learn, TensorFlow
Machine Learning Algorithms
- Logistic Regression, Random Forests, Naive Bayes, KNN, XGBoost
Evaluation & Performance metrics
- Decision Matrix, feature importance
- pyenv with Python: 3.8.5
Use the requirements file in this repo to create a new environment.
pip install -r requirements.txt