Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/analitico-771/machine_learning_trading_bot

This is an Application that implements an algorithmic trading strategy that uses machine learning to automate the trade decisions
https://github.com/analitico-771/machine_learning_trading_bot

financial-analysis hvplot logistic-regression machine-learning moving-average pandas-dataframe predictive-modeling python scikit-learn stock-price-prediction support-vector-machine

Last synced: 26 days ago
JSON representation

This is an Application that implements an algorithmic trading strategy that uses machine learning to automate the trade decisions

Awesome Lists containing this project

README 

        



[![Contributors][contributors-shield]][contributors-url]

[![Forks][forks-shield]][forks-url]

[![Stargazers][stars-shield]][stars-url]

[![Issues][issues-shield]][issues-url]

[![LinkedIn][linkedin-shield]][linkedin-url]



  Table of Contents

  
    

        
  1. 

          About The Project

          

          

        
    2. 

        
  2. 

          Evaluation Report

        
    3. 

        
  3. 

          Getting Started

          

        
    4. 

        
  4. Usage
    5. 

        
  5. Roadmap
    6. 

        
  6. Contributing
    7. 

    	

        
  7. Contact
    8. 

        
  8. Acknowledgements
    9. 

      



#

#### *** Disclaimer! This application is for educational purposes only and not intended for financial advice.  The dataset used is historical OHLC price data from an emerging market asset.  The symbol of the asset is not disclosed due to this app being developed strictly for educational purposes only ***

#

## About The Project



This is an Application that implements an algorithmic trading strategy that uses machine learning to automate the trade decisions.



##### There's an issue with rendering on GitHub - use the link below to see the project

##### http://nbviewer.org/github/AnaIitico/machine_learning_trading_bot/blob/main/machine_learning_trading_bot.ipynb



### Built With



* [Python](https://www.python.org/)

* [Python conda](https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html)

* [Python JupyterLab](https://jupyter.org/)

* [Python CSV Reading/Writing](https://docs.python.org/3/library/csv.html)

* [Python pandas](https://pandas.pydata.org/)

* [Python scikit-learn](https://scikit-learn.org/stable/)



## Evaluation Report



#### Overview of the Analysis



* I built 3 models.  First I established a `Baseline Strategy`, then build a model based on `Support Vector Machine Strategy` method and then a `LogisticRegression Strategy` model.

* The purpose of the analysis is to train a model that uses machine learning to automate the trade decisions.

* The dataset is historical OHLC price data from an emerging market asset.  The symbol of the asset is not disclosed due to this app being developed for educational purposes.

* The variables that is being predicted is the `signals_df['Signal']` column which is calculated based on the cross of two moving averages used as a trigger.  The moving averages are assigned to the `X` variable : `X = signals_df[['SMA_Fast', 'SMA_Slow']]` and the `y` variable is assigned the value of : `signals_df['Signal']` for the model.

* Rules :

  - When the `SMA_Slow` crosses the `SMA_Fast` from a lower to a higher price it sends a `1` to the `y` `(signal)` variable

  - Wen the `SMA_Slow` crosses the `SMA_Fast` from a higher to a lower price it sends a `-1` to the `y` `(signal)` variable

  - A `1`=`Buy` and `-1`=`Sell`.

* Assumption - FULL DISCLOSURE - This model assumes that the trader is able to consistently enter and exit the market at the previous day close whenever the `Signal` is generated.  This is a critical and dangerous assumption since it assumes 100% accuracy and 100% success in the price of the market execution for the `Actual Returns` calculation.  A more realistic and accurate entry strategy will be needed in order to fine tune a profit strategy, but this was beyond the scope of this analysis. Understanding the risks and assumptions of the model proves this to be a good model since the machine learning aspect of the analysis was proven and it is profitable.



#### Summary of Analysis



* The `Baseline Strategy` performance didn't yield good results considering that it is predicting an outcome solely based on a cross of two moving averages that have been shifted forward by one period (one day in this model) to predict a future crossing point.

![Baseline Screenshot](images/baseline_strategy_returns.png)



* The `Support Vector Machine Strategy` performance was very good overall considering the response during periods on volatility in the market.  The  `Strategy Returns` column yielded better results thant the `Actual Returns` column for this strategy almost consistently. This model attempts to establish a relationship based on geometrical properties of data. It tries to finds the distance that separates the distance between the classes - in this case the classes are the classifications into the `signals_df['Signal']` column as a `1` or `-1`.

![Baseline Screenshot](images/support_vector_machine.png)

* Changing the `SMA_Slow` to a larger number decreased the overall returns as shown in the `SMA_Slow = 13` screenshot where you can see that the `Actual Returns` line matches the `Strategy Returns` line.

![Baseline Screenshot_13](images/support_vector_machine_13.png)



* The `LogisticRegression Strategy` performance was worse in comparison and seemed to stop working after some time. It attempts to establish a dependent type of relationship between the two moving averages which is not the best approach for this model.  Predicting a dependency between the `SMA_Slow` and the `SMA_Fast` creates a fragile relationship vulnerable to overfitting, which the model exposed.

![LogisticRegression Screenshot](images/logisticregression_strategy_returns.png)

* Changing the `SMA_Slow` to a larger number increased the overall performance slightly but still has catastrophic results in the end.

![LogisticRegression Screenshot_13](images/logisticregression_strategy_returns_13.png)



* RECOMMENDATION - The `Support Vector Machine Strategy` is the best strategy for this model and asset.  Markets have geometrical properties of data, as well as unstructured and semi-structured properties of data and this is the best approach for the strategy of predicting two crossing moving averages. 



## Getting Started



* Install and import the required libraries and dependencies ONLY as needed!



* You don't need Python installed in your computer. You can install Anaconda and JupyterLab normally just like any other application on your computer. Follow the instructions for Anaconda, ensure that its working, then install JupyterLab.



* I have placed Comments throughout the code so that you can follow the code and be able to replicate the app on your own. Also, so that you're able to contribute in the future :-)



### Prerequisites



A text editor such as [VS Code](https://code.visualstudio.com/) or [Sublime Text](https://www.sublimetext.com/)



### Installation



1. Clone the repo

   ```sh

   git clone https://github.com/AnaIitico/machine_learning_trading_bot.git

   ```



2. You don't need to install pip - Conda comes with pip and you can also use the command

    conda install 'package name'

   

3. Install Conda according to the instructions based on your operating system.

    For windows users you MUST use the Administrator PowerShell. Users with AMD Processors MUST use the Administrator PowerShell 7 (X64) version

  

    Once installed Conda has an Admin PowerShell version shortcut - look on your Start menu for it.

    This shortcut will prove very useful at times when you need to install other apps or make adjustments to your installation



    Once installed you will see (base) on your terminal

   

4. Activate Conda Dev environment

   ```sh

   conda activate dev

   ```

   You should now see (dev) on your terminal



5. Install JupyterLabs

   ```sh

   pip install jupyterlab

   ```



6. Run JupyterLabs

   ```sh

   jupyter lab

   ```

   A browser window should open on localhost:8888/lab



## Usage

  



## Roadmap



  The app is finished



See the [open issues](https://github.com/AnaIitico/machine_learning_trading_bot/issues) for a list of proposed features (and known issues).



## Contributing



Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are **greatly appreciated**.



1. Fork the Project

2. Create your Feature Branch (`git checkout -b feature/AmazingFeature`)

3. Commit your Changes (`git commit -m 'Add some AmazingFeature'`)

4. Push to the Branch (`git push origin feature/AmazingFeature`)

5. Open a Pull Request



## Contact



Jose Tollinchi - [@josetollinchi][linkedin-url] - [email protected]



Project Link: [https://github.com/AnaIitico/machine_learning_trading_bot](https://github.com/AnaIitico/machine_learning_trading_bot)



## Acknowledgements

#### Other Dependencies used to build the project.

##### Search google for the correct conda install command as needed

* numpy

* hvplot

* matplotlib



#### Other Acknowledgements



* [Img Shields](https://shields.io)

* [Choose an Open Source License](https://choosealicense.com)



[contributors-shield]: https://img.shields.io/github/contributors/AnaIitico/machine_learning_trading_bot.svg?style=for-the-badge

[contributors-url]: https://github.com/AnaIitico/machine_learning_trading_bot/graphs/contributors

[forks-shield]: https://img.shields.io/github/forks/AnaIitico/machine_learning_trading_bot.svg?style=for-the-badge

[forks-url]: https://github.com/AnaIitico/machine_learning_trading_bot/network/members

[stars-shield]: https://img.shields.io/github/stars/AnaIitico/machine_learning_trading_bot.svg?style=for-the-badge

[stars-url]: https://github.com/AnaIitico/machine_learning_trading_bot/stargazers

[issues-shield]: https://img.shields.io/github/issues/AnaIitico/machine_learning_trading_bot/network/members?style=for-the-badge

[issues-url]: https://github.com/AnaIitico/machine_learning_trading_bot/issues



[linkedin-shield]: https://img.shields.io/badge/-LinkedIn-black.svg?style=for-the-badge&logo=linkedin&colorB=555

[linkedin-url]: https://www.linkedin.com/in/josetollinchi/