I'm relatively new to ML but my goal it to use Tensorflow.js and build a ML model that can help me detect a certain wave formation for an automated trading system. Examples of the 3-leg pattern I am trying to identify:

enter image description here

I've been studying ML concepts but I still can't wrap my head around how to start writing actual code in tensorflow.js in relation to this problem, what would be the logical steps to create this model? Can someone point me in the right direction, perhaps even show some untested code that might help me understand how to go about this problem.

  • $\begingroup$ There is a lot going on with the graphs that you have included, which aspects of the chart are you trying to predict? (black line, pink line, both pink lines?) $\endgroup$
    – Ethan
    Mar 3 '19 at 23:40
  • $\begingroup$ The black lines indicating a 3 leg pattern is was I am trying to identify. I currently have coded logic that can identify it based on (Bill Williams) fractals and refined with other methods like checking for sufficient retracements of the 2nd leg, however due to the high degree variation of these patterns I think it may be a better task for ML, just not sure how to do it. $\endgroup$
    – parliament
    Mar 4 '19 at 12:53
  • $\begingroup$ Welcome to the site! Can you share a little more about what your source data is going to look like? Those charts can be created with any number of data sets and it really won't be possible to help you unless we know that the source data is like. $\endgroup$ Mar 4 '19 at 14:16
  • $\begingroup$ Follow-up, have you ever worked with Tensorflow or writing LSTM network architecture before? $\endgroup$
    – Ethan
    Mar 4 '19 at 17:03
  • $\begingroup$ @I_Play_With_Data The source data is standard candlesticks with timestamp, volume, and asset prices for open, high, low, and close. $\endgroup$
    – parliament
    Mar 4 '19 at 17:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.