As we mentioned a few weeks ago, here we are with some more details about our first international installation at the end of January in the metropolitan area of Jersey City (USA).
As a result of the last few years of research and development we have installed our automated Trading solution that implements the latest machine learning techniques for time series forecasting.
The 1.0 product uses a combination of different algorithms including SVM (support vector machine), Random Forest, Decision Tree for sentiment analysis and < strong> combination of a neural network MLP (multi layer perceptron) as a regressor connected to a pattern matching system for the estimation of future prices and the calculation of the probabilities of success.
This decision-making engine is actually connected to Metaquote’s Metatrader platform (in fact the standard for the retailer trader), from which it receives real-time data that is used for continuous updating and improvement of the strategy.
The bridge provided for the connection with MT4 also allows the user to define his / her risk profile, the use of a money management stage and the setting of some parameters useful for the management of tolerances for the execution of orders due to latency and spred imposed by the broker.
Software was built using machine learning libraries like Tensorflow, Keras and Scikit-Learn.
Currently the system is only tested on the Foreign Exchange Market, commonly known as Forex,
in its major currencies AUD, CAD, CHF, EUR, GBP, NZD, JPY and in the crosses derived from them as it uses
of the typical and exclusive characteristics of this type of market.
The currently released version, has 2 years of track records on real trading accounts with globally certified brokers and has shown a success rate in execution of orders equal to 85% with peaks up to 91%.
The maximum drawdown recorded is 9.8% (another parameter configurable via the bridge on MT4) with
an average ROI that stands at 4.5% monthly. (I’ve been low 😊 !!!)
In the next article we will discuss what are the future scenarios and the research and development areas where M.IT is moving towards the next version of our software.