Quote from: legendster on August 06, 2019, 06:22:48 PM

Please share some relevant data. You will find that many of us experienced traders here use chart patterns, and technical indicators to calculate and predict market movement fairly accurately. Can you, or rather would you be willing to share which TI's you are using here?

Binance have API for data in history monly and daily, we crawl it and build a Model for machine learning learn history data, and use the term Neural network to speak about machine learning, you only need to understand that a neural network is a kind of a machine learning technique and Long Short Time Memory(LSTM) is a type of neural network (cf LSTM ).

LSTM
LSTM for Long Short Time Memory is a model design that processes past data to give a prediction. Its a type of neural network.
Lets illustrate this with an example.
If I tell you give me the next number after 6, what can you say ? Naturally you will give 7 because it is the next one, but why not 8 if the series is N+2 ?
Ok, now I tell you give the next number after 3, 6, and now you can answer with 9 or 12, because the link between 3 and 6 can be N+3 or N*2. Then I tell you give me the next number after 0,3,6 and you see that the answer is N +3.
https://miro.medium.com/max/700/1*o-OCmzzqCZvY6GEI4BNDBg@2x.jpeg

Here, you implemented, in your brain, a simple series check. You used the past values to predict the next one. However, you need enough values in the example range to be able to predict the right result, or at least the most probable one.
Your brain is able to make that for a large variety of data. For example, when you look at a ball thrown in the air, you can predict where it will fall, because your brain analyses the position at different steps, and can predict the next steps until the final position.
With LSTM, it is a bit more complicated but the basics are the same:
we send a chronological series to let the neural network find the next probable value. A huge dataset, build on a long period, is not needed. You need enough data, to split the dataset in small chronological chunks, for the neural network to be able to understand and predict the next value.

For more detail, you can read the full articles at this link: https://medium.com/smileinnovation/how-to-make-profits-in-cryptocurrency-trading-with-machine-learning-edb7ea33cee4

We based on that articles for build own project

Thanks