This repo contain codes to predict future stock price by machine learning algorithms.

A series of ML models are implemented.

• Random Forest
• AdaBoost
• XGBoost
• DNNs
• CNNs
• RNNs
• Transformer

• Python 3.6
• Scikit-Learn 0.20
• XGBoost
• PyTorch1.0

Due to credential reason, raw dataset cannot be publicised for now. Processed data may be uploaded later.

Raw data is first processed via Data_Preprocess.ipynb, which is fundamental step for whole project.

• For those ensemble models, simply run corresponding jupyter notebooks cell by cell

• For NN models, go to ./Regression_NN run Training.py

  Training.py has a series of parameter you can adjust. For the details you may need to read the Training.py source code.

  Here is an example for training Transformer model:

  python Training.py --batch_size 1024 --prev_step 30  --suffix pre30_class_raw --num_classes 1 --epochs 30 --feature_size 256 --normalize --layers 1  --model Transformer

Codes are in one of my partners' repo