📈 Temporal Financial Forecasting with RNNs, GRUs, and LSTMs

This project implements a deep learning pipeline for multi-horizon prediction of financial time series using temporal neural architectures. The framework supports multivariate inputs and outputs, enabling robust forecasting of market signals like Open, High, Low, Close, and Volume.

---

🚀 Key Features

• 📊 Multivariate Time Series Forecasting (OHLCV)
• 🔁 Multi-step Prediction Window: Past 15 days → Next 3 days
• 🧠 Models: RNN, LSTM, GRU (stacked, configurable)
• 🔍 Hyperparameter Grid Search over:
  • Learning Rates
  • Batch Sizes
  • Number of Layers
  • Hidden Units
  • Optimizer Selection (Adam, SGD)
• ✅ Evaluation via RMSE and visual diagnostics
• 📉 Real-time plots:
  • Training/Validation loss trends
  • Predicted vs Actual market movement
  • Architecture comparisons by RMSE

---

🧰 Use Cases

• Financial modeling & algorithmic trading research
• Signal forecasting in time series (economics, IoT, crypto)
• Sequence learning & multivariate regression frameworks

---

📊 Benchmark Summary

Model	Best RMSE	Optimizer	Learning Rate	Batch Size	Layers	Hidden Units
GRU	0.024	Adam	0.001	16	2	100
LSTM	0.026	Adam	0.01	32	2	50
RNN	0.028	Adam	0.01	64	2	50

✅ GRU-based models demonstrated the strongest performance in multi-horizon OHLCV forecasting.

---

📈 Visualizations

• 🔹 RMSE bar charts by model and configuration
• 🔹 Best predicted vs actual plots (OHLCV)
• 🔹 Loss curves per epoch
• 🔹 Future market trend simulations