💳 FinFlow AI – Payment Risk Prediction Dashboard

FinFlow AI is a real-world fintech project that predicts transaction failures, detects payment frauds, and intelligently routes to the most reliable payment gateway — all in real-time using machine learning.

🔍 What It Does:

• Predicts failure probability based on gateway, latency, retry count, etc.
• Flags suspicious/fraudulent transactions
• Recommends the best payment gateway (based on live performance metrics)

💻 Tech Stack:

• Python – Core logic & scripting
• Scikit-learn – ML classification models
• Pandas, NumPy – Data wrangling
• Streamlit – Dashboard UI

DEMO LIVE (https://finflow-payment-ai-in.streamlit.app/)

🔧 Project Structure

finflow_payment_ai/ │ ├── data/ │ ├── payments_dataset_sample_5k.csv # Raw sample data │ └── cleaned_payments_dataset.csv # Cleaned after ETL │ ├── models/ │ ├── failure_predictor.pkl # Failure model │ ├── fraud_detector.pkl # Fraud model │ ├── encoder.pkl # Encoder for failure model │ └── fraud_encoder.pkl # Encoder for fraud model │ ├── scripts/ │ ├── etl_pipeline.py # Data cleaning + feature engg │ ├── model_train_failure.py # Train failure model │ └── model_train_fraud.py # Train fraud model │ ├── dashboard/ │ └── app.py # Streamlit frontend app │ ├── utils/ │ └── helpers.py # Optional utils (load_pickle etc.) │ ├── requirements.txt # All dependencies └── README.

---

⚙️ Setup Instructions (Run Step by Step)

✅ Step 1: Clone and enter project

git clone https://github.com/prashanthsword/finflow_payment_ai.git cd finflow_payment_ai python -m venv venv

Activate:

Windows

venv\Scripts\activate

Mac/Linux

source venv/bin/activate

✅ Step 2: Install required libraries

pip install -r requirements.txt

✅ Step 3: Run ETL to clean the raw data

python scripts/etl_pipeline.py

✅ Step 4: Train both models (failure + fraud)

python scripts/model_train_failure.py python scripts/model_train_fraud.py ✅ Step 5: Launch the Streamlit Dashboard

streamlit run dashboard/app.py 📈 Dashboard Features 💥 Predicts failure probability of a transaction

⚠️ Detects fraud risk

📌 Suggests the most reliable gateway (based on failure rate + latency)

🎛️ Input options: Amount, Gateway, Payment Type, Device, Latency, etc.

🧠 Realtime ML prediction using trained models

About This Project :

This was a solo-built ML dashboard inspired by real-world fintech systems. It combines backend data pipelines, model training, and UI in one complete pipeline — ideal for showcasing end-to-end ML engineering skills.