jojoBot


[OverView](/jojoBot/docs/MD/OverView.html) | [DOW Stable Base Line](/jojoBot/docs/MD/StableBasdelineDowJones.html) | [Trading Experiments](/jojoBot/docs/MD/READMExperiment.html) | [PaperTrading](/jojoBot/docs/MD/READMExpAlpacaPaperTrading.html) | [TECH](/jojoBot/docs/MD/README.TECH.html) | [FAQ](/jojoBot/docs/MD/READMEfaq.html) | [SnapShot](/jojoBot/docs/MD/READMECodeSnapShot.html)

Renforced Learning Trading Bot


AI StockBot

JojoFin with the three layers: market environments, agents, and applications. For a trading task (on the top), an agent (in the middle) interacts with a market environment (at the bottom), making sequential decisions.

[Introduction](OVERVIEW) | [SnapShot](READMECodeSnapShot.md) | [PaperTrading](READMExpAlpacaPaperTrading.md) | [FAQ](READMEfaq.md)

Introduction

If you’re tired staring at charts all day, and doing trades manually while exploring daily market data just relax and let the bot do all the hard work. This Trading-bot operates 24/7, ensuring no trading opportunities are missed. An AI JOJO Trading Bot offers the benefits of automated trading without needing constant attention, giving regular investors a chance for higher returns. The name of the AI trading bot is JojoFin. It is like having an automatic helper that trades for you 24/7 based on set rules, quickly making lots of small trades to profit from market changes, while traditional investing involves buying assets and holding them for a long time with less frequent trades and lower risk.

Reinforcement Learning (RL)

Reinforcement Learning (RL) techniques are considered convenient for this task : In this experiment, we present an RL environment for the portfolio optimization based on state-of-the-art mathematical formulations. The environment aims to be easy-to-use, very customizable, and have integrations with modern RL frameworks. Jojobot is a library that will allow you to easily create trading robots that are profitable in many different asset classes, including Stocks, Options, Futures, FOREX, and more. Check your trading strategies against historical data to make sure they are profitable before you invest in them. JojoBot makes it easy for you to do (backtest) your trading strategies and easily convert them to algorithmic trading robots.

Experiment Overview

  1. Pull 1 year of trading data for (Insert your stock, options or crypto) with Yahoo Finance Downloader API
  2. Create a simulated trading environment using real trading data.
  3. Train an neural network to predict that Stock Price using reinforcement learning inside this simulation with FinRL
  4. Once trained, backtest the predictions on the past 30 days data to compute potential returns with FinRL
  5. If the expectd returns are above a certain threshold, buy, else hold. If they’re below a certain threshold, sell. (using Alpaca API)

In order to have this to run automatically once a day, we can deploy it to a hosting platform like Vercel with a seperate file that repeatedly executes it.

Dependencies

from config file, TRAIN , TEST and TRADE days

TRAIN_START_DATE = '2010-01-01'
TRAIN_END_DATE = '2021-10-01'
TRADE_START_DATE = '2021-10-01'
TRADE_END_DATE = '2023-03-01'

Yahoo donloader for data frames collection from Start Train to End Tradedate

df = YahooDownloader(start_date = TRAIN_START_DATE,
                      end_date = TRADE_END_DATE,
                      ticker_list = config_tickers.DOW_30_TICKER).fetch_data()

Features Included DOW_30_TICKER - Technical, VIX and Turbelance INDICATORS,

fe = FeatureEngineer(
                      use_technical_indicator=True,
                      tech_indicator_list = INDICATORS,
                      use_vix=True,
                      use_turbulence=True,
                      user_defined_feature = False)

Envionment Aeguments

 env_kwargs = {
      "hmax": 100,
      "initial_amount": 1000000,
      "num_stock_shares": num_stock_shares,
      "buy_cost_pct": buy_cost_list,
      "sell_cost_pct": sell_cost_list,
      "state_space": state_space,
      "stock_dim": stock_dimension,
      "tech_indicator_list": INDICATORS,
      "action_space": stock_dimension,
      "reward_scaling": 1e-4
  }

Taining Agents Ensamble

  
  models = {
      "a2c": trained_a2c,
      "ddpg": trained_ddpg,
      "ppo": trained_ppo,
      "td3": trained_td3,
      "sac": trained_sac
  }

  results = predict_with_models(models, e_trade_gym)
  # Access results for each model
  df_account_value_a2c = results["a2c"]["account_value"]
  df_account_value_ddpg = results["ddpg"]["account_value"]
  df_account_value_ppo = results["ppo"]["account_value"]
  df_account_value_td3 = results["td3"]["account_value"]
  df_account_value_sac = results["sac"]["account_value"]
  #### Taining Agents Ensamble

predict_with_models Ensamble

  def predict_with_models(models, environment):
      for model_name, trained_model in models.items():
          df_account_value, df_actions = DRLAgent.DRL_prediction(
              model=trained_model,
              environment=environment
          )
          results[model_name] = {
              "account_value": df_account_value,
              "actions": df_actions
          }

      return results

Tutorial

Google Colab Notebooks

Examples for Stocks, Options, and Crypto in the notebooks provided below. Open them in Google Colab to jumpstart your journey!

Notebooks Open in Google Colab
Stocks Orders Open In Colab
Options Orders Open In Colab
Crypto Orders Open In Colab
Stock Trading Colab

Features

This site is open source. Improve this page.