An archive of the models from the article. You can train your own, because the models can be very different, depending on the training hyperparameters.
Machine learning models are trained in Python and converted to ONNX format. You don't need to change their parameters and code in the encludnik.
The settings are simple:
sinput double MaximumRisk=0.001; //Progressive lot coefficient sinput double ManualLot=0; //Fixed lot sinput ulong OrderMagic = 666; //Orders magic input int max_orders = 1; //Orders number input int stoploss = 2000; //Stop loss input int takeprofit = 2000; //Take profit input string comment = "The ONNX EA";
But you can add your own filters to try to make the models trade better.
At least you can optimise stoploss and take profit, on any timeframe
The model is connected via a header file, e.g.:
#include <EURUSD ONNX include propensity matching original IPW.mqh>
31 March 2024. Added examples from the article about "clustering".
"ONNX Trader Clusters" bot, ONNX models and include file. Clustering is done by k-means method.
Training is done from 2010 to 2020. Forward - from 2020 to 2024.
The"Onnx Trader" bot supports the second method of training from the article"Matching trades using clustering"
To do this, connect the iniclude file <EURUSD ONNX include10.mqh>, which will pull up the corresponding models "catmodel10.onnx" and "catmodel_m10.onnx"
Models are trained on EURUSD H1, but can be used on any timeframe!
