Subscribe to the Alpha Quant Program (using this link www.quantreo.com) before the Black Friday (Friday 29th November) to obtain our "Machine Learning for Trading" course for FREE !
Oh finally. Someone who isn't Moon Dev who actually knows what they're talking about! We don't got enough people like you on this website so I appreciate you making these videos. Also, Python is very based.
@@TechThroughHistory thanks a lot for you support. If you liked it, you should definitely think about joining the AQP before Black Friday before it is less than 5% of all we will see on the « Machine Learning for Trading » course I will release! Have an excellent day! I’m still here if you have any questions!
Learning Python myself so I can eventually get into cybersecurity, though for going for data analytics atm. Personal recommendations for job boards or companies to apply too?
I want to ask you , What do you think about rolling normalization , as the scaling the whole data set at once lead to information leakage by default , that breach the fundamental principle of CPCV test , because we using the same distribution of overall dataset for each spit.
Hi, it is indeed a very good question. 1) Here if you fit your StandardScaler class on the X_train ONLY and you transform your X_test using it, there is not any leakage value. The Data leakage come when you fit and transform on all the features and then separate into X_train and X_test. 2) However, it can be interesting some time to use a rolling normalization especially for trending features (prices for example) but the problem is that you will at this end not have a really standardize features in a lot of case which is the purpose of the StandardScaler. So, you have other advantages and weaknesses and you need to do your choice taking them all into account +the problem you are solving and the features you are using. By the way, I do not really see the link with the CPCV as the data leakage in the CPCV is removed by removing the beginning and the end of each set and then apply the Standardization if needed, so it is even more restrictive than what I did here as I didn't remove for example the first rows of the test set. Let me know if you have any other questions! Have an excellent day!
Subscribe to the Alpha Quant Program (using this link www.quantreo.com) before the Black Friday (Friday 29th November) to obtain our "Machine Learning for Trading" course for FREE !
Oh finally. Someone who isn't Moon Dev who actually knows what they're talking about! We don't got enough people like you on this website so I appreciate you making these videos. Also, Python is very based.
@@TechThroughHistory thanks a lot for you support. If you liked it, you should definitely think about joining the AQP before Black Friday before it is less than 5% of all we will see on the « Machine Learning for Trading » course I will release! Have an excellent day! I’m still here if you have any questions!
Learning Python myself so I can eventually get into cybersecurity, though for going for data analytics atm. Personal recommendations for job boards or companies to apply too?
@@TechThroughHistory Unfortunately not for cybersecurity sorry. Have a nice day!
😂😂😂
I want to ask you , What do you think about rolling normalization , as the scaling the whole data set at once lead to information leakage by default , that breach the fundamental principle of CPCV test , because we using the same distribution of overall dataset for each spit.
Hi, it is indeed a very good question.
1) Here if you fit your StandardScaler class on the X_train ONLY and you transform your X_test using it, there is not any leakage value. The Data leakage come when you fit and transform on all the features and then separate into X_train and X_test.
2) However, it can be interesting some time to use a rolling normalization especially for trending features (prices for example) but the problem is that you will at this end not have a really standardize features in a lot of case which is the purpose of the StandardScaler. So, you have other advantages and weaknesses and you need to do your choice taking them all into account +the problem you are solving and the features you are using.
By the way, I do not really see the link with the CPCV as the data leakage in the CPCV is removed by removing the beginning and the end of each set and then apply the Standardization if needed, so it is even more restrictive than what I did here as I didn't remove for example the first rows of the test set.
Let me know if you have any other questions!
Have an excellent day!
Hello Lucas , how to subscribe alpha quant Program , I couldn't see the link in discription
@@majdsaud8123 here is the link to join the Alpha Quant Program www.quantreo.com! Let me know if you have any questions!