/Transformer-for-Stock-Price-Prediction

Primary LanguageJupyter NotebookMIT LicenseMIT

Transformer-for-Stock-Price-Prediction and Golden-Cross Strategy

  • ZHANG MIAN
  • CentraleSupĂ©lec

This file is a brief introduction about Transformer Model and the application in finance domain.

What we have got

The prediction of the Moving Average (10days)

image

The loss function with the epochs.

image

1. Introduction to Transformer Model

The fundamental thesis of Transformer Model: Link: https://arxiv.org/pdf/1706.03762.pdf

2. Import and labs and preprocess the data

In this project we calculate the percentage return of the daily stock price(Open, Close, High, Low, Volume)

3. Construct Transformer model by Tensorflow.

  • Time Embedding
  • Single Head Attention
  • Multi Head Attention

4. Moving Average Case and Golden-Cross Strategy

Here we define the moving average of 20 trading days, all rest the same. And also we recalculate the 60 days moving average. If the golden-cross shows up, we decide to buy the stock(Also this should be conducted with RSI or other index).