I found it’s hard to study since I do not know every code what is it
is it better to skip the data part and study it later when I need
what Is the hands-on metho for this book
I had the same experience. I read the entire book with limited retention. Then I went back and did the hands-on. I found that I could spend months on one chapter or concept applying it to my strategies.
1 Like
I would say the book is not designed for beginners. It helps a lot to have something in mind that you wish to build. I highly recommend teaming with a few programmers to build something together, as a group. It will shorten the learning curve.
1 Like
Hi Blake, I noticed you mentioned this book is not for beginners. I have understood Python’s basics but haven’t learned anything about ML yet. Just recently I got exposed to automated trading through trading simulation software. What kind of resources/platform you recommend me to go through first to close the gap before tackling the book?
My technique, make so many mistakes that you need to go back to the relevant chapters to see why your ML models is making a zillion dollars. But, I have to say that after a year of studying, with literally no prior experience even in python, I’ve learned a lot. Recently, I bought a Packt video course from Udemy. I figured since this books was from Packt it should be of decent quality. Right away, I figured out something was amiss. Despite the video series being called Machine Learning for Finance, it was committing one of the most basic mistakes for financial time series, lookahead bias. The course was supposedly “predicting” forex/stock prices, but it was predicting on using features of current MA or RSI, which you already need the close price to know. Then, even though the lecturer already acknowledges the temporal component of financial time series, he doesn’t split his data to recognize the concept. Anyway, point being, I went from almost zero knowledge of programming and Machine Learning to being able to spot these flaws. Mostly, through making mistakes with lookahead bias, overfitting, etc and then referring back to this book to see what I’m doing wrong.
But ok, my suggestion on how to study this book. Focus on a model, such as binary classification and lightgbm, and just try to follow the entire workflow. Finding data, feature engineering, hypertuning the model, predicting, and then really see if you can implement it in real trading. I think that the only part I think this book misses, is that big step towards final production.