Home Community Crypto-ML Roadmap, Feature, and Suggestions Details about market data and ML?

Viewing 1 reply thread
  • Author
    Posts
    • #9619
      achapela
      Participant

      Hello,

      I read the FAQs and everything I could find about your trading ML. I could not find information about what are the market data you are using, more specifically the ones that have the most weight. I also am curious about the actual ML implementation beyond the expression “Deep Neural Network”. Do you use a classic Multi-layer Perceptron, a Convolutional Neural Net, a Long Short Term Memory…? Maybe some reinforcement learning? A mix of those?

      Without giving away your recipe, I am curious to hear more. One reason for me to be curious is that I recently spent a lot of time just looking at charts on tradingview.com and buy doing so and using very common indicators I was able to make gain everyday trading simply by hand. So I would hope the ML to way out perform me because it can learn about all the correlations (especially time correlation) and also recognize OHLC patterns (like an ascending triangle, ascending channel…).

      Thank in advance!
      A.

    • #9624
      Justin
      Moderator

      Hey, great questions!

      Probably the thing that will give you the most directional insight into our approach is here: https://crypto-ml.com/blog/bitcoin-price-prediction-with-diy-machine-learning-in-excel/

      While it’s considerably more basic than what we run, a lot of the core concepts are discussed.

      In terms of the actual ML implementation, the answer is it depends–we can run many configurations and let it determine the best. The configuration will change over time.

      Specifically, we use these: https://bigml.com/api/deepnets

      And yes, our concept is very similar to what you described in your TradingView experience. We seek to provide the ML all of the building blocks for technical trading. It should be able to setup a variety of equations from them and thereby determine market movement.

      To me, that’s the most interesting part of an ML project…how have you structured the inputs?

Viewing 1 reply thread
  • You must be logged in to reply to this topic.