GIVEN-CHINYAMA/usd-zmw-exchange-rate-forecasting

Forecasting the USD/ZMW exchange rate using ARIMA & LSTM deep learning | Stationarity testing · ACF/PACF analysis · Model Comparison · 12-month forecast | Python · Statsmodels · TensorFlow · Kwame Nkrumah University # 📈 Forecasting the USD/ZMW Exchange Rate Using ARIMA and LSTM Models **Author:** Given Chinyama  |  **Institution:** Kwame Nkrumah University  |  **Date:** May 2026 --- ## 📌 Project Overview This project forecasts the USD/ZMW (US Dollar to Zambian Kwacha) exchange rate using two time series forecasting models — ARIMA (a classical statistical model) and LSTM (a deep learning model). The goal is to compare both approaches and determine which one better captures the trends and patterns in the exchange rate data. --- ## 🎯 Objectives - Collect and preprocess historical USD/ZMW exchange rate data - Perform Exploratory Data Analysis (EDA) to understand trends and patterns - Build and evaluate an ARIMA model for time series forecasting - Build and evaluate an LSTM (Long Short-Term Memory) neural network model - Compare the performance of both models using error metrics - Visualize forecast results against actual exchange rate values --- ## 📊 Model Results | Model | MAPE | Accuracy | |-------|------|----------| | ARIMA(3,1,2) | 20.70% | 79.3% | | **LSTM** | **3.57%** | **96.4%** | > 🏆 **LSTM significantly outperformed ARIMA** — achieving 96.4% forecast accuracy vs 79.3% for ARIMA on 2,608 daily observations (2015–2024). --- ## 🛠️ Tools & Technologies | Tool | Purpose | |------|---------| | Python | Core programming language | | Google Colab | Development environment | | Pandas | Data manipulation | | NumPy | Numerical computation | | Matplotlib / Seaborn | Data visualization | | Statsmodels | ARIMA model | | TensorFlow / Keras | LSTM model | | Scikit-learn | Data preprocessing & evaluation | --- ## 📂 Project Structure ``` usd-zmw-exchange-rate-forecasting/ │ ├── usd_zmw_exchange_rate_forecasting.ipynb # Main notebook ├── README.md # Project documentation ├── LICENSE # MIT License └── .gitignore # Python gitignore ``` --- ## 📊 Models Used …