rennykefs/MPESA-Liquidity-API

A forecasting tool to help agents predict their daily liquidity needs, optimizing their cash and float balances. # M-Pesa Liquidity Forecasting System **An End-to-End AI Solution for Mobile Money Agent Optimization** ### Project Overview This system predicts daily cash requirements for M-Pesa agents in Kenya. Using **SARIMAX (Seasonal Auto-Regressive Integrated Moving Average with Exogenous Factors)**, it accounts for payday cycles, holidays, and school fee windows to prevent "stock-outs" and optimize cash-in-transit routes. ### Architecture - **Backend:** FastAPI (Python) - High-performance API serving models from RAM. - **Frontend:** Streamlit - Interactive dashboard for bank operations managers. - **Model:** SARIMAX time-series forecasting with `joblib` serialization. ### Quick Start 1. **Clone the repo:** `git clone github.com` 2. **Install dependencies:** `pip install -r requirements.txt` 3. **Run the API:** `uvicorn Scripts.main:app --reload` 4. **Run the Dashboard:** `streamlit run dashboard.py` ### Features - **Real-time Prediction:** Get instant "Withdraw/Deposit" advice for specific agents. - **Network Summary:** High-level view of liquidity risk across all agent categories (Urban/Rural). - **Dynamic Scaling:** Automated model loading using Python Lifespan events.