Predicting stock prices with LSTM network

This project goal is to demonstrate how to use LSTM networks and apply in some real data.

Getting Started

In this repository, you will find the following notebook:

• Moving Average: First look at the data, a SMA and EMA model
• Sanity Check: A test with LSTM and the model used to make sure the model works
• LSTM input and output: An explanation on how I manipulated the data to use as input/output for the model
• LSTM Single Company: A LSTM model used to predict the closing price of a single company. Note that the notebooks ‘Single Company B’, ‘Single Company C’ and ‘Single Company D’ are extremely similar.
• LSTM multi-companies prediction: A LSTM model used to predict the closing price of all four companies (A, B, C and D).

Prerequisites

The code was written in Python, and uses:

• Pandas and Numpy (for data manipulation)
• Matplotlib (for data visualisation)
• Keras (for building and training the model)
• sklearn (for data scaling)

Results

A LSTM network is not able to predict the closing price behaviour by only looking at its historical data.