Swedish

Screenshots

Technical Analysis with Neural Networks

Neural networks are perfectly suited to model price movements. They can model price behavior mathematically themselves and they have the ability to extract information from large amount of data which is necessary for complex signals such as financial price movements.

A neural network is made up of a number of interconnected neurons which behave like an artificial brain. The network is stimulated by appropriate input signals, in this case of historical price data and outputs from other technical indicators. The neural network is trained to find connections between these input signals and future price levels. If there exists such a connection, neural networks have an amazing capability to find these. This is then used to generate appropriate trading signals.

Neural network settings in Optimal Trader »»»

 

The biological brain

The humane brain is one of the most complex items in the universe. Although the brain has a considerably slower processing speed and less storage capability than a modern computer, it is the nevertheless superior in many areas such as pattern recognition and adaption to new situations. With modern technology neural networks can imitate the brain's way to learn and process information. Neural networks are basically a collection of interconnected neurons, each one with several inputs and one output. The inputs' weight (importance) and numbers can vary. The output is a function of all inputs. The picture below shows how biological neurons interact with each other.

We see that the blue neuron is sending an impulse to the yellow neuron. The yellow neuron may receive other impulses (varying in strength) from other neurons, but is only sending one signal (a function of all the inputs).

This biological system can be transfered to a computer:

The diagram to the right shows a number of inputs with varying strength which the neuron processes and produces an output from. The building of networks of these neurons is a lot more complicated than described here and involves neural models and mathematics that are beyond this short summary.

Although neural networks are not nearly as sophisticated as their biological counterparts, they mimic our way to think to make quicker and better decisions.

 

Trading with neural networks

• Neurala networks do not follow any predefined rules (as classic technical analysis does) but finds its own connections and rules.
• Neural networks are highly accomplished in finding patterns
• Neural networks can work with different types of inputs and are therefore more flexible than a single indicator. You can feed a neural networks with both historical data and with different technical indicators.
  
  

Who Uses Neural Trading Systems ?

There are various financial areas in which financial professionals use neural-based applications: in various bond trading to determine the probability that a given company will default on its debt, in sales forecasting to predict future sales based on a series of inputs, and in equity trading to predict future price movements. Many professionals and institutions are testing and using this technology to more accurately predict events and profit.

Why Doesn't Everybody Use It?

The application of neural networks to trading is relatively new and build neural applications involves complex mathematics. You cannot analyze a neural network and explain how a specific network draws its conclusions.

Some critics say that the idea of a software system learning the behavior of the market is flawed. If humans cannot predict the market with certainty, how can we create software that can achieve something we do not even fully understand?

Academic studies

There are many empirical studies performed on colleges and universities which show that neural networks can be used in order to yield higher profit than a simple "buy and hold" strategy. Please read more in the collection below of technical dissertations:

Stock Market Prediction Using Artificial Neural Networks
Birgul Egeli, Meltem Ozturan, Bertan Badur
Bogazici University, Istanbul, Turkey

Testing Stock Market Efficiency Using Neural Networks
Marius Januskevicius
Stockholm School of Economics in Riga, Lithuania

Neural Network Applications in Stock Market Predictions
Marijana Zekic
University of Josip Juraj Strossmayer in Osijek, Croatia

Feedforward and Recurrent Neural Networks and Genetic Programs for Stock Market and
Time Series Forecasting
Peter C. McCluskey
Brown University, Rhode Island

Using Neural Networks to Forecast Stock Market Prices
Ramon Lawrence
University of Manitoba

Regime Switching and Artificial Neural Network Forecasting of the Cyprus Stock Exchange Daily Returns
Eleni Constantinou, Robert Georgiades, Avo Kazandjian and Georgios Kouretas
University of Crete, Greece

Stock Price Prediction: Kohonen Versus Backpropagation
Alexey Zorin
Technical University of Riga, Lithuania

Stock Prediction – A Neural Network Approach
Karl Nygren
Royal Institute of Technology, Stockholm

Modelling Riga Stock Exchange Index Using Neural Networks
Alexey Zorin and Arkady Borisov
Technical University of Riga, Lithuania

Hurst Exponent and Financial Market Predictability
Bo Quian and Khaled Rasheed
University of Georgia, USA
(Experiments with neural networks showing that series with large Hurst Exponents can be predicted more accurately than those series with Hurst Exponents close to 0.5.)