AI Trading Assistant now It’s Open Source
I called this AI Netrade because it’s based on neural network to recognize cryptocurrency or stock chart and candle stick pattern. I develop this model for almost 1 month doing try and error from model version 1 to model version 2, and after I develop and tested it the model is able to make profit for me, I get profit about 3% after I run this model for almost 4 days. Let’s dive in
The story behind it
I am a cryptocurrency traders, couple months ago I bought bitcoin at 5 PM and at 9 PM I fall as sleep plus I loss momentum to sell my bitcoin then I loss about -5% it’s make me unhappy.
Because of that I want an assistant who works for me 24 hours a day and 30 days in a month without stop. Beside cryptocurrency trader I am a software engineer and I am artificial intelligence practitioner, I believe artificial intelligence is about teaching computers about human.
How Netrade Works
Netrade is neural network specifically convolutional neural networks that able to recognize which chart and candle stick pattern that will bring the price go up or go down this model did it very well. Let’s dive into how this model works.
This model takes input chart pattern image and candle stick pattern image it’s looks like this
I made 2 blocks convolutional layers for each image (chart pattern and candlestick ) and this model produce 2 output (0 price will go down and 1 price will go up). Here is the architecture looks like
The performance for this model is extremely satisfying as shown in graph below
Result Profit and Loss
What I write here is netrade version 2.0.0, we tested it for almost 4 days and we give this model investment capital about IDR Rp. 1,000,000 ( one million rupiah ) and we get return about 3% as shown from this graph
We trigger this model to predict the pattern every 15 minutes, we still get minus because of we need more data to train this model. We train this model with 450 chart pattern images and 450 candle stick pattern images. And here is the pie chart about win and loss rate
Before this version
Before this version we made 3 models that we wrote in this article if you would to read you can read that here
Now it’s open source
If you want to change or improve it you can do it now. I believe together is better and I know all of you is better than me, let me learn from you.
