Member-only story
How to Improve a Machine Learning Model’s Trading Strategy
Automatically enhance trading backtests using parameter tuning
If you have been reading along with my previous articles, you can probably see that I have been hard at work trying to build a machine learning powered trading strategy. The goal of this strategy is to beat the market without much input from myself (besides the initial code development), which is easier said than done. Many of my previous attempts have centered around backtesting a trading strategy that utilizes a specific time-series model called Facebook Prophet. This model’s strategy was usually backtested against the cryptocurrency market or together with sentiment analysis.
This time around I’ll be backtesting it against the good-old stock market. The goal of this trading strategy is to simply outperform a basic buy and hold strategy of the same stock and still end up profitable. But this time, I’ll be enhancing it by tuning the many custom parameters I made for this AI powered trading strategy. Often times I have found that these parameters need to be adjusted differently for different stocks. However, I shouldn’t need to manually change them every time I backtest a new stock.
In order to automate this task, I’ll need to accomplish the following:
