Text Size
Friday, August 23, 2019

Gandalf Store




Artificial Intelligence

Artificial intelligence and the most modern technologies of genetic engineering in the service of trading...

Artificial intelligence is the branch of engineering that deals with replicating intelligent behavior in machines. The world of computing has drawn from this framework for different application fields such as optimization of the development of electronic circuits, marketing solutions and forecasting of complex processes. The first application in the financial sector dates back to the eighties, when several research groups tried to predict the course of stock prices using neural networks. 

Genetic Algorithms, Neural Networks and Fuzzy Logic as tools for development of Trading Strategies.

Since the early results, partly because of the changing nature of the market, it became clear that the winning element was the use of genetic algorithms to the elaboration of strategies fully automated.

The idea is both simple and extremely powerful: generate a population of artificial investors who based their behavior on the markets on a genetic mapping, covering the largest number of possible rules. This algorithm is used both to decide the input position and the closure of the trade. Subsequently, the population is measured on the reference market, on the basis of a cost function that determines the priority. The n best investors are then saved, that mutual exchange part of their rules, creating a new generation of investors. The process is repeated a large number of times and enriched by multiple algorithms of evolution that have the task of overcoming suboptimal solutions.

The result of this methodology is a set of investors who not only have the best performance on the training period, but they have guaranteed reliable statistics even in areas not known in the data set.

But what exactly does it mean to discover profitable rules? Imagine creating a genetic map with all the possible relationships between data prices of the last N bars (for instance the last 6 bars). You will obtain a matrix of rules similar to the one in the picture below:

Imagine managing a matrix of thousands of elements. Now if you let the machine define random rules inside this matrix, you will obtain a certain number of activations. In this example you would have obtained:




In general the machine will build a pattern that could consist of price, volume, cyclic or indicator rules. The same procedure will define the exit rules. After discovering the enter and the exit rules, the syntetic investor obtained in this way will develop a position management criteria. For example it will control the position with a static monetary stop, a dynamic stop like a chandelier stop, a profit target or a time exit.

Now, if this set of environments will generate a substantial profit or if it will survive in the ranking of best syntetic investors for the selected cost function, we will obtain a new trading system candidate.

This process is one of the beating hearts of G.A.N.D.A.L.F. (Genetic Algorithms Network for Discover Adaptive Laws in Finance).

Once that a set of syntetic investors have been integrated in an organic trading system, we use a portfolio set of rules to decide how long to keep a single investor and in which way to substitute it.


Privacy & Cookies Policy

Questo sito utilizza i Cookies per migliorare la navigazione. Utilizzando questo sito e continuando nella navigazione si intende accettata la Privacy Policy e la Cookies Policy. Puoi bloccare in ogni momento questa raccolta di informazioni seguendo le istruzioni contenute in queste pagine.


Intelligenza Artificiale

L' Intelligenza Artificiale e le più moderne tecnologie di Ingengneria Genetica al servizio del Trading...



Gandalf Project è sia un Laboratorio di Finanza Quantitativa che una Società di Consulenza...



La scelta di Python come codice elementare per la modellizzazione di ambienti complessi...