FOSSONG/ai-production-forecasting

AI-driven production forecasting using LSTM, Prophet, and Random Forest ensemble for Niger Delta oil fields # AI-Driven Production Forecasting Using LSTM, Prophet, and Random Forest ## Overview This repository contains the complete implementation of an AI-driven ensemble framework for oil production forecasting, integrating Long Short-Term Memory (LSTM), Prophet, and Random Forest (RF) models. The workflow is developed and validated using 32 years (1992–2024) of monthly production data from the Gabo Field, Niger Delta. The study demonstrates how ensemble learning improves forecast robustness and accuracy relative to traditional Decline Curve Analysis (DCA) and standalone machine learning models, particularly in heterogeneous reservoir settings. ## Methodology The forecasting framework consists of: - Individual base learners: LSTM, Prophet, and Random Forest - Two ensemble strategies: - Performance-weighted averaging - XGBoost-based stacking - Evaluation using MAE, RMSE, and Mean Absolute Scaled Error (MASE) against a naïve baseline The methodology is designed to be transparent, reproducible, and computationally efficient for operational deployment.