Algorithmic Trading Bot using Machine Learning

Have you ever wanted to invest and make tons of money without wanting to put in the time and effort to trade? What if I said you could...

Comment 1

Inspiration

Ever since middle school, I have been interested in finance and computer science. High school gave me the opportunity to intertwine both of my passions together. At HackPHS 2022, I gained the inspiration to build my very first algorithmic trading bot.

What it does

This trading bot trades for me without me actually having to make a trade. To come up with an initial pool of stocks to invest in, I analyzed the top 10 stocks out of each sector to see which ones produced the most reward with the least risk. Once, I came up with a pool of stocks, I narrowed them down together so that they were optimized to their best ability. I then implemented a pairs trading algorithm that takes two pairs of stocks and hedges them. Each paired stock shares a negative to none correlation with its other stock. As one stock goes down, another stock is needed to balance out the loss, so that's how the pair works. After I did all of the research, I implemented the modeling onto QuantConnect's platform where I simulated the model on historical data(backtesting).

How we built it

I built it using Python 3.10.2 and Python on QuantConnect's software. I also used several of Python's libraries such as pandas, numpy, scikit-learn, statsmodels, scipy, matplotlib, and seaborn.

Challenges we ran into

I ran into a major challenge in using the QuantConnect software. I developed my Pairs Trading Model and Reward-to-Risk Ratio asset analyzer and I needed to implement it into the backtesting software. This backtesting software tests the algorithm on historical data. QuantConnect's documentation was something that I was unfamiliar with and caused me a difficult time. I overcame this setback by reanalyzing the stack each time.

Accomplishments that we're proud of

I developed a trading bot with a Sharpe Ratio of 1.683. The Sharpe Ratio is a metric that is used to measure the return on investment given the risk. The S&P 500 index's Sharpe Ratio is -0.78 and anything between 1 to 3 is said to be strong. I also implemented a Monte Carlo Simulation in my research where I analyzed the relative value at risk and relative shortfall.

What we learned

Through this project, I was able to learn and become familiar with quant connect's software. I also learned a lot about K-means Clustering and its implementation in finance. The most important thing that I took away from this project is to constantly think like a computer if an error is present in the code.

What's next for Algorithmic Trading Bot using Machine Learning

The next part of this is to develop new trading bots with higher Sharpe ratios and larger expected returns. I would also like to continue expanding my knowledge of the machine learning techniques that are present in this field of quantitative finance/financial engineering.

Built With

Share this project:

Updates