NanLi-1217/Cocoa-Commodity-Volatility-Forecasting

Time-series forecasting (ARIMAX) and risk modeling (GARCH) of global cocoa prices using Ghana climate data. # Cocoa Commodity Price Forecasting & Volatility Analysis 🍫📈 ## 📌 Project Overview This project introduces statistical modeling methods to predict daily cocoa prices based on time series and regression models, focusing on the influence of climatic conditions in Ghana on market behavior. The project employs historical cocoa price data from the International Cocoa Organization and daily climatic data from Ghana. The overall objective is to compare a range of modeling frameworks (ARIMA, ARIMAX, and GARCH) to ascertain which offers the best balance of accuracy, interpretability, and forecasting capability. ## 🛠️ Tech Stack & Methods * **Language:** R (forecast, rugarch, tseries, dplyr, ggplot2, car) * **Time Series Modeling:** ARIMA, ARIMAX (with exogenous climate regressors) * **Volatility Modeling:** GARCH(1,1) with Student's t-distribution * **Statistical Diagnostics:** Augmented Dickey-Fuller (ADF) test, Ljung-Box test, Variance Inflation Factor (VIF), Cross-Correlation Function (CCF) ## 📊 Key Business Insights & Results ### 1. Climate-Driven Forecasting (ARIMAX vs. ARIMA) Proved that external climate factors (e.g., lagged precipitation and temperature) act as leading indicators for cocoa price shocks. By incorporating these exogenous variables, the ARIMAX model improved the AIC by over 12,500 units compared to the univariate baseline, successfully capturing price spikes that traditional models missed. This provides actionable signals for proactive commodity procurement. ### 2. Market Risk & Volatility (GARCH) Fitted a GARCH(1,1) model to quantify conditional variance, revealing a strong volatility persistence ($\alpha + \beta = 0.96$). This confirms a "long memory" in market shocks, providing a critical quantitative foundation for options pricing and risk hedging strategies in highly volatile commodity markets. ## 📂 Repository Structure * data: Contains aggregated historical cocoa prices (ICCO) and Ghana climate data (NOAA). * cript: Modularized R scripts …