Neural Rabbit
- 专家
- Evgeniy Scherbina
- 版本: 1.2
- 激活: 20
The advisor "Neural Rabbit" trades simultaneously several symbols in a fully automated mode: AUDUSD, EURUSD, GBPUSD, USDCAD, USDCHF, and USDJPY.
I am getting the first results with this signal:
https://www.mql5.com/en/signals/2079266/
The advisor uses a recurrent neural network to analyze data from different timeframes. It uses prices variations and values of indicators as inputs.
The main advantage of this advisor is the use of normalized data in the machine learning. This approach to training neural networks has allowed the improvement of both accuracy and loss to spread over 200 epochs. Verification was done using consecutive bars of 3 months. This is very important. Amateurs usually conduct a machine learning with the verification by a certain amount of random bars. However, consecutive sequences of bars are very much similar. Grouping of bars by 3 months has helped avoid repetitions in bar sequences. Therefore, my approach to training of this advisor has allowed to efficiently search for pattern-like moves of the markets without naive curve-fitting.
A huge amount of tests I did allowed to make one powerfull conclusion. A machine learning aimed at improving the forecast accuracy learns the history too fast. It actually learns every nervous tick of the market. Its accuracy in the unknown period may also be high, but, unfortunately, its loss is also very high. If a strategy makes 7 good decisions out of 10, it is 70% of accuracy. Good? Bad! It is bad if the loss is high because any bad forecast may be so bad that it will eat away all hopes for profitability. Therefore, it is the strategy with training by loss that is capable to handle well the unknown period, even though its history result may be thinly over the satisfactory level.
Start the advisor in one chart. Regardless of your current chart, the advisor always trades all its symbols at a time.
You can start the advisor "Neural Rabbit" on a chart of any symbol and any timeframe.
Recommended settings for live trading
- Min deposit 250 USD (1000 USD is better)
- Close on signal change = True
- Training type = By_Loss
- Session start (hour) = 4
- Trail = 100
- Volume = 0.01% (for USD deposits)
Properties
- Suffix >> it can be handy for those who have different symbols with different suffixes in their terminal. Most users can leave this field empty.
- Close on signal change >> tries to close a trade without a trail, when its signal changes to opposite.
- Training type >> Training by accuracy creates an illusion of reliability because it rushes through the history so nicely (over-fitting). Training by loss learns major moves of the market, but not all of its nervous bursts. Therefore, it is the strategy with training by loss that should be preferred for trading in the future, unknown period.
- Session start (hour) >> set the hour to start a trading session from. New trades will be open starting with this hour. This may help avoid gaps at the start of the day.
- Takeprofit
- Trail
- Stoploss
- Volume >> either a fixed volume per trade (ex: 0.02) or a percent of available equity (ex: 0.015%). The advisor prints the actual volume in the "Experts" tab upon start or timeframe change. If your account does not use USD as the deposit currency (that is, euro, yuan, AUD, etc.), make sure it calculates a volume correctly or use a fixed volume.
