Gas Price Predictor

Built for an AI class at the University of Houston, as part of a 4-person team comparing different machine learning approaches to price forecasting. I built one of the project’s LSTM models and contributed to data collection and research. The team trained and compared LSTM, GRU, and Random Forest models using technical indicators (moving averages, RSI, Bollinger Bands, MACD) derived from historical USOIL prices, plus a separate Random Forest model predicting weekly Houston gas price changes from crude and regional market data. GRU outperformed LSTM on this dataset (lower error), though both models struggled to anticipate sudden price spikes driven by real-world events outside the dataset, such as wars and policy shifts, a key finding of the project.