This project implements a complete deep learning pipeline for forecasting complex, multivariate time-series data using a stacked LSTM neural network. The dataset is programmatically generated to simulate realistic non-stationary behavior with seasonality, trend, and noise—similar to energy consumption or financial market data.