The use of neuro networks for financial series prediction
Currently, we observe the explosion of interest to neural networks in different field. Neural networks are a
powerful tool for forecasting nonlinear dependencies. Because neural networks are widely used for financial time series predicting. To solve
a problem, where linear approximation working poor, the use of linear models
are not entirely correct. Neural networks are based on primitive biological
model of nervous systems. The concept arose in the study of processes in the
brain while thinking, and when trying to simulate these processes.
The use of financial series prediction.
I would like to note that prediction is possible only when the future to some extent depends on the past. Unable to create Grail for trading on the forex market with the help of neural networks. But to improve the current trading system by properly setting objectives and the preparation of the data will turn out.
1y Stage - Choice network topology
To predict the financial series most applicable multilayered pertseptron or network Vorda, because expert assessment already laid in the original data and can be allocated among them.
Pic. 1 Multilayered Pertseptron
2y Phase Selection network characteristics
For pertseptron we can selected number of layers.
3rd Phase Selection of learning parameters
4y Stage Network training
We can split the data sampling on learning sampling and test sampling, which checks learning outcomes.
If performance on the test sample that improved, education can be considered successful.
In our opinion, the most promising is the development of networks to complete the existing trading systems.
let us suppose, if we have a trading system has been successful in the trend phases of forex market (for the trend part most applicable linear model) and impressions on Drawdown on flat part.
In this case, develop a network to flat part of forex market should complement and improve existing trend system.
Flat -A price that is neither rising nor declining.
Trend - The general direction of a market or of the price of an asset. Trends can vary in length from short, to intermediate, to long term. If you can identify a trend, it can be highly profitable, because you will be able to trade with the trend.
BJF Trading Group
http://iticsoftware.com
Forex Fully Automated Trading System |
Metatrader Expert Advisor Forexy GBPUSD Independent Monitoring |
 |
The use of financial series prediction.
I would like to note that prediction is possible only when the future to some extent depends on the past. Unable to create Grail for trading on the forex market with the help of neural networks. But to improve the current trading system by properly setting objectives and the preparation of the data will turn out.
1y Stage - Choice network topology
To predict the financial series most applicable multilayered pertseptron or network Vorda, because expert assessment already laid in the original data and can be allocated among them.
Pic. 1 Multilayered Pertseptron
2y Phase Selection network characteristics
For pertseptron we can selected number of layers.
3rd Phase Selection of learning parameters
4y Stage Network training
We can split the data sampling on learning sampling and test sampling, which checks learning outcomes.
If performance on the test sample that improved, education can be considered successful.
In our opinion, the most promising is the development of networks to complete the existing trading systems.
let us suppose, if we have a trading system has been successful in the trend phases of forex market (for the trend part most applicable linear model) and impressions on Drawdown on flat part.
In this case, develop a network to flat part of forex market should complement and improve existing trend system.
Flat -A price that is neither rising nor declining.
Trend - The general direction of a market or of the price of an asset. Trends can vary in length from short, to intermediate, to long term. If you can identify a trend, it can be highly profitable, because you will be able to trade with the trend.
BJF Trading Group
http://iticsoftware.com
Comments