/AutoStockTrading

Auto stock Trading Models for DS340W at Penn State

Primary LanguageJupyter Notebook

DS340w Project - Automatic Stock Trading

  • This is a project repository for DS340w course.

  • The project is focusing on stock price prediction.

  • The project is highly influenced by the previous research - Deep Reinforcement Learning for Automated Stock Trading: An Ensemble Strategy (https://github.com/AI4Finance-Foundation/FinRL/blob/master/tutorials/2-Advance/FinRL_Ensemble_StockTrading_ICAIF_2020.ipynb)

  • The name scheme of each jupyter notebook is:

    *** TrainingStartYear_Novelty_VersionOrdinal *** (VersionOrdinal means the order we created each version)

    • several different novelty
      • 6indicators --- adding 2 additional tech indicators (features) to the original model (4 tech indicators already included in the origibal model)
      • 5Agents --- 5 Agents ensembled (PPO, A2C, SAC, TD3, DDPG)
      • Valid --- change of validation ratio
      • NULL --- no novelty
      • NewAgent --- replace one of the orginal agents that was ensmbled in parent paper

  • The structure of this repository:

    • 'ParentPaper.pdf' ----- pdf of parent paper

    • 'Original_2009_ParentPaper.ipynb' ----- implementation of original parent paper (data cover upto 2021-07-06)

      • Sharpe Ratio 1.96

    • 'Original_2009_Updated_ParentPaper.ipynb' ----- updated implementation of original parent paper (data cover upto 2022-03-21)

      • Sharpe Ratio -0.14

    • 'FinalModel_2006_NULL_version3_2.ipynb' ----- The final version of modified implementation -

      • Sharpe Ratio 0.94

    • 'FunVersion_2006_NULL_version3_2.ipynb' ----- The final version of modified implementation with our chosen stocks

      • Sharpe Ratio 0.64

    • '340Paper.doc' ----- Final Paper for DS340W

    • SingleAgentModel ----- Folders of 5 single agent model

      • '2006_A2C.ipynb' ----- Sharpe Ratio 0.57
      • '2006_PPO.ipynb' ----- Sharpe Ratio 0.58
      • '2006_SAC.ipynb' ----- Sharpe Ratio 0.16
      • '2006_TD3.ipynb' ----- Sharpe Ratio 0.58
      • '2006_ddpg.ipynb' ----- Sharpe Ratio 0.44

    • FinRL-Library ----- Folders of FinRL library, which is created by the authors of parent paper

      • 2 modification made by us under (finrl --- drl_agents --- stablebaselines3)
        • 'models.py' --- original ensemble method created by authors of the library (A2C, PPO, and ddpg)
        • 'models_modified.py' --- ensemble A2C, PPO, and TD3
        • 'models_modified2.py' --- ensemble A2C, PPO, SAC, TD3, and ddpg

    • 2009Model ----- data start traing year in 2009

      • '2009_6indicators_version2.ipynb' ----- Sharpe Ratio 0.86

    • 2006Model ----- data start traing year in 2006

      • '2006_5Agents_version8_3.ipynb' ----- Sharpe Ratio 0.36
      • '2006_6indicators_version3.ipynb' ----- Sharpe Ratio 0.57

    • 1998Model ----- data start traing year in 1998

      • '1998_6indicators_version4.ipynb' ----- Sharpe Ratio 0.51
      • '1998_NULL_version5.ipynb' ----- Sharpe Ratio 0.97
      • '1998_NewAgent_version6.ipynb ' ----- Sharpe Ratio 0.56
      • '1998_Valid_version5_2.ipynb' ----- Sharpe Ratio 0.81

    • 1997Model ----- data start traing year in 1997

      • '1997_5Agents_version8_2.ipynb' ----- Sharpe Ratio 0.57

    • 1962Model ----- data start traing year in 1962

      • '1962_5Agents_version8.ipynb' ----- Sharpe Ratio 0.19
      • '1962_6Indicators_version7.ipynb' ----- Sharpe Ratio 0.20