Fraud Detection Kaggle Project

This repository demonstrates a practice project in the fraud detection field using data from Kaggle competition - IEEE-CIS Fraud Detection.

File Documentation

This project consists of 2 major parts: the fraud detection analysis notebook and the dash app.

dash folder: this folder contains the python scripts and assets for the dash app.

notebooks folder: this folder contains two notebooks EDA.ipynb and Fraud_Detection.ipynb and the Fraud_Detection.ipynb notebook contains the end to end fraud detection analysis.

Fraud Detection Analysis

The fraud detection analysis contains 4 sections:

• Data Cleaning;
• Feature Engineering;
  • Missing values analysis
  • Dimensionality Reduction
    1. for categorical variables, merge small levels
    2. for numerical variables, PCA
• Fraud Detection Modelling
  • lightGBM;
  • XGBoost;
  • Hyperparameter fine-tuning + cross validation
• Feature Mmportance Analysis.

Dash App Screenshot

The dash app is coded in python and two screenshots are shown as follows.