"Security is not just an ideal, it is a necessity." – Eleanor Roosevelt
| Date | Version | Team | Organization | Description |
|---|---|---|---|---|
| 13/12/2024 | 01 | Riskova | No Country | MVP |
| Item | Descripción |
|---|---|
| Team | Riskova |
| Project | Predictive Fraud Detector (PFD) |
| Start Date | 11/11/2024 |
| End Date | 13/12/2024 |
| Client | NoCountry |
| Project Leader | Hernán Casasola |
| Project Manager | Jose Ibarra |
The Riskova project aims to create a platform to assist in detecting fraud in electronic payments, utilizing machine learning techniques and data analysis. The dataset used for this project was sourced from Kaggle. https://www.kaggle.com/datasets/kartik2112/fraud-detection
Develop a predictive model to detect fraudulent transactions on an electronic payment platform using machine learning techniques and behavioral analysis.
- Exploratory Data Analysis: Perform interactive data analysis to uncover insights such as distributions, correlations, and outliers.
- Data Preprocessing: Prepare the data by cleaning, normalizing, encoding, and scaling, ensuring readiness for model training.
- Model Training: Develop and train predictive models to detect fraudulent transactions, leveraging machine learning techniques.
- Real-Time Fraud Detectio: Simulate real-time fraud detection with transactional data, showcasing system performance.
- User Interface: Provide an interactive web-based interface using Streamlit for seamless user interaction and data visualization.
- Report Generation: Generate downloadable reports with visualizations and key metrics for comprehensive analysis summaries
- Python Version: Ensure Python 3.x is installed.
- Dependencies: Required Python libraries must be installed (listed in
requirements.txt). - Streamlit: Streamlit should be installed to run the web application interface.
- Data Files: CSV files to be processed
| Item | Description |
|---|---|
| Home | Introduction to the credit card fraud problem, its significance, and the areas involved. Navigation to key system sections. |
| Data Ingestion & Exploration | Upload datasets in CSV format or connect to databases. Interactive data exploration: distributions, correlations, and outliers. |
| Data Preprocessing | Data cleaning and transformation: handling missing values, duplicates, and scaling. User-customizable transformations. |
| BI Dashboard | Interactive dashboard for key KPIs like fraud percentage, trends, and geographic analysis. Dynamic charts and visualizations. |
| Predictive Modeling | Model training for fraud prediction. Evaluation metrics such as precision, recall, and F1-score. User testing of hyperparameters. |
| Advanced Analysis | Model interpretation using SHAP. Fraudulent behavior segmentation with clustering techniques. |
| Real-Time Detection | Real-time prediction simulation with uploaded or queried data. Visualization of predictions and associated explanations. |
| Reporting & Export | Generate downloadable reports in PDF or Excel format with key metrics and charts. Includes an executive summary of the analysis. |
| Documentation | Technical details of the project, tools used, and team roles. System usage guide. |
| Step | Command | Description |
|---|---|---|
| Clone the project | https://github.com/No-Country-simulation/c22-31-m-data-bi.git | Clone the project repository to your local machine. |
| Install dependencies | pip install -r requirements.txt | Install all required dependencies for the project. |
| Run Predictive Fraud Detector | streamlit run main.py | Run the app with Streamlit |
You can try the live demo by clicking on the following link:
