This repository holds code for the following multi-step-ahead prediction methods: autoregressive integrated moving average (ARIMA) model, stacked RNN, stacked LSTM network, and stacked GRU. Each method's performance is measured in terms of prediction accuracy using MBE, MAPE, and RMSE, as well as average computational run-time. The methods are trained on the univariate time series of global horizontal irradiance (GHI) from satellite-measured data across varying locations throughout the United States. We predict GHI for five years using both short and long-term prediction intervals.

Project for forecasting Tesla (TSLA) stock prices using advanced LSTM neural networks. Includes a single-step ahead model and a multi-step stacked LSTM model for short and medium-term predictions. Data-driven insights for stock market enthusiasts and practitioners.

Project for forecasting Tesla (TSLA) stock prices using advanced LSTM neural networks. Includes a single-step ahead model and a multi-step stacked LSTM model for short and medium-term predictions. Data-driven insights for stock market enthusiasts and practitioners.

No description available

Trains stacked bidirectional multi-step LSTM on XRPUSDT OHLC data to predict high volatility buy/sell zones

Multi-step time-series forecasting on the Jena Climate dataset. Stacked LSTM trained from scratch in PyTorch, served via FastAPI, tracked with MLflow, and deployed as a live Gradio demo — type any city and get a 48-hour forecast.