Stock Price Prediction with Keras and Jupyter Notebooks

Overview

This project is a simple introduction to Jupyter Notebooks, TensorFlow/Keras, and time series forecasting. The goal is to use a neural network to predict stock prices based on historical data.

Current Findings

• The Keras neural network currently underestimates total price and volatility.
• Predictions tend to be reactive rather than proactive—when a sudden drop occurs, the model responds rather than predicting it beforehand.
• Further tuning and optimization are needed to improve forecasting accuracy.

Next Steps

• Learn more about hyperparameter tuning and time series forecasting techniques.
• Experiment with LSTMs, GRUs, and attention mechanisms to improve performance.
• Optimize feature engineering and training methodology for better predictive capability.