Dedalo314/Stock-prediction-using-transformers

Code to train stock prediction models based on a combination of convolutional neural networks and transformers.

Stock prediction using transformers

In this repository some models are included to perform stock prediction with deep learning. The best and final version is Orpheus5Classifier, which can be trained easily with PyTorch Lightning in one GPU with 2GB of memory.

Orpheus5Classifier is trained to predict if a company's stock is going to increase or decrease, given the previous 2 years of stock prices. The final accuracy of Orpheus5Classifier over any company of the NASDAQ index is around 60%.

NOTE: These models are not intended for real investment, only as an experiment to test the potential of deep learning in the prediction of stock markets.