Random Forest Stock Trading Model

This project investigates the in-sample and out-of-sample trading performance of a manual strategy and a Random Forest classification-based learner, both built using three technical indicators:

• Bollinger Bands
• Relative Strength Index (RSI)
• Stochastic Oscillator

In both the manual strategy and Random Forest learner, we used the same starting cash of $100,000. Allowable positions are 1000 shares long, 1000 shares short, or 0 shares. We can trade up to 2000 shares at a time, with unlimited leverage, assum- ing commission of $9.95 and trade impact of 0.005 (0.5%). JPM (JPMorgan Chase & Co.) is the only symbol used. The in-sample time period is from January 1, 2008 to December 31, 2009, and the out-of-sample period is from January 1, 2010 to December 31, 2011.

Subsequently, we also explore the effects of different values of market impact (0.0005, 0.005, 0.05) on cumulative returns and the number of trades of the Random Forest learner during the in-sample time period.

---

Instructions to Run the Project Locally

Prerequisites

Ensure you have Python and the following libraries installed:

• NumPy
• Pandas
• Matplotlib
• SciPy

Alternatively, if you have Anaconda/Miniconda installed:

1. Create the environment with the provided environment.yml file:

   conda env create --file environment.yml

2. Activate the environment:

   conda activate ml4t
   

Running the Project

1. Excecute the project script:

   python testproject.py
   

Viewing Generated Outputs

Performance charts: Located in the images folder In-Sample performance statistics: Performance - In-Sample for JPM.txt Out-of-Sample performance statistics: Performance - Out-of-Sample for JPM.txt Market impact on in-sample performance: Market Impact - In-Sample for JPM.txt