Trading Signals

AI & Machine learning
1

How do I extract trading signals from the SVM strategy?  I want to evaluate my strategy against future data.  



Thanks,

J

2

Hello John, thanks for your query! Well, there are many ways you can do that. 



The idea is that you need to have a set of features for which you'll label the position you take. For example, one such feature can be the difference between the daily open and close. 

 

AAPL['Open-Close'] = AAPL.Open - AAPL.Close

Here you're creating a new feature from the AAPL daily opening and closing prices.

Now once we have this, we can very easily reckon the signal. Just make sure the Df['Close']  of a certain day is greater than the Df['Close'] of the day before. Here is how that can be done, 
 
y = np.where(Df['Close'].shift(-1) > Df['Close'],1,-1)

Here, we're creating a new vector where we put in 1 if the close of a given day is greater than the close of the previous day ( because you'd want your algorithm to buy it) and -1 if it's not. 

So, for the SVM, AAPL['Open-Close']  will be the predictor(X) and the vector y will be the target variable. 

Of course, there are many other ways the signal can be generated. Do get back, if you'd like to discuss further! 
 
3

Dear Sir,



I was about to ask the same question and found John's post here. I have a question regarding your answer:



Should we generate the labels for the future? for example after generating the target variable by your method should we do y.shift(-1) so we use the feature for today to predict what will happen tomorrow?



Thanks in advance,

4

Hi Mohammed



If you check the code above for "y", that is what we are in fact doing. We compare tomorrow's price (Df['Close'].shift(-1)) with today's price (Df['Close']) and give a "Buy" signal, for today as per the condition.



Therefore, we need not separately do "y.shift(-1)" again. 



Hope that answers your query.











 

5

Yes got it now. Thanks a lot!