Exclusive Interview | NeuroShell for Financial Forecasting
NeuroShell uses neural networks to analyze your favorite indicators, recognize multi-dimensional patterns too complex to visualize, predict and forecast market movements and then generate trading signals based upon those patterns, predictions and forecasts. The CEO of NeuroShell spoke with us on how traders can use machine learning and AI to help find profitable trading rules.
Q: Hi Denham, thank you for taking the time to speak with us today. I wanted to introduce you to the QuantNews audience so that people can learn more about your unique product. Your website mentions that your company has been involved in artificial intelligence, neural networks and genetic algorithms since the 1980s. Can you tell us more about your company?
A: Ward Systems Group began in 1988 when our founder and my father, Steve Ward, admired the prediction and classification capabilities of neural network models, but thought there had to be a better way to implement these techniques for common modeling problems. Until then, neural network software was pretty much in the realm of academia and computer programmers. He developed the original NeuroShell software that let the user concentrate on using neural networks to quickly solve problems in a well-designed user interface, but without having to learn computer programming or become a neural network expert.
Q: What kind of individuals or industries have you used your product?
A: Our neural net software products are used by everyone from individuals and small businesses to large Fortune 500 companies and government agencies. Businesses use our general purpose software for sales forecasts, inventory predictions, estimating transaction volume and utility load estimates. Medical applications include disease diagnosis, patient scheduling, workload prediction, and monitoring physiological signals. Scientific applications include spectral analysis and interpretation, oil exploration, quality control and physical system modeling.
Of course, NeuroShell Trader is targeted towards individual and institutional traders interested in predicting price movements of stocks, futures, exchange rates, etc. or discovering multidimensional market patterns and trends too complex to be visualized.
NeuroShell Trader is targeted towards individual and institutional traders interested in predicting price movements of stocks, futures, exchange rates…
Q: Can you tell us more about how you got started designing technology for financial markets applications?
A: After several years in the neural network business, Ward Systems Group noticed that the majority of sales and tech support questions for our general purpose neural network products were increasingly slanted towards financial applications. That’s when I began to develop a product custom tailored for financial markets and the result was NeuroShell Trader
Q: For somebody who is interested in using machine-learning technology or artificial intelligence (AI) for trading, but perhaps not a programmer, how would they use your service?
A: NeuroShell Trader was designed for traders, not programmers. The user interface is wizard based and takes you step by step through indicator design, building trading rules and constructing neural network predictions. Trading systems and predictions can be easily applied to multiple securities. Intermarket analysis is easily accomplished by adding data streams from different markets into the trading rules or predictions.
All of this and a lot more is done without having to write a single line of code or script. This is why NeuroShell Trader has become one of the top systems used by traders.
This is why NeuroShell Trader has become one of the top systems used by traders.
Q: Your platform has two extremely interesting bits of functionality. The user is given the ability to create a ‘Prediction’ as well as a ‘Trading Strategy.’ Could you take a moment to describe their differences from a developer sense?
A: Trading Strategies are rule based trading systems in which you supply the rules. One of the most well-known and simplest examples is a moving average crossover system which buys and sells as different sized moving averages cross over each other. While rule based trading is pretty common in most other trading software, what differentiates NeuroShell Trader is the ability to create complex custom trading systems without having to write any code or scripts.
NeuroShell Trader then goes beyond standard technical analysis systems by adding neural network predictions. Neural networks simulate how our human brain builds neuron connections as we learn and experience new things in life. Take for example a young child that begins seeing more and more animals. Their brain learns to differentiate between a house cat, tiger and lion, even when they see a new one that is slightly different than the others they saw before. In the case of trading, the neural network is likewise learning to recognize complex market patterns using past historical data and it can then recognize similar patterns in the future.
A good rule of thumb is if you know the rules for your trading system, then build a trading strategy. However if you don’t know what trading rules to use or want to create a unique system outside of traditional rules being traded by others, then you will want to create a neural network prediction using a platform such as NeuroShell Trader.
(A picture of the NeuroShell platform, Image courtesy of WardSystems)
Q: For a trader who may not be fully involved with the actual coding design of the algorithmic trading strategy, what practices or ways can people use your platform to help analyze performance/risk? In other words, what functionality is available for a client to gauge the risk of trade automation?
A: Like most trading software products, NeuroShell Trader includes the standard trading statistics report that shows trades, profit/loss, drawdown, etc. But then it goes much further since those statistics alone aren’t enough to help the trader fully evaluate the risks of automating trading.
The greatest risk occurs in overfitting a trading system. Overfitting occurs when you try to increase the profitability of your system by tweaking the rules yourself a few too many times or when you let the computer over optimize the system. The result is a trading system that looks great in hindsight, but loses money in the future once you begin automated trading.
To evaluate how much overfitting has occurred and thus the risk of automated trading, NeuroShell Trader includes three unique features: out of sample evaluation, paper trading optimization and walk forward optimization. All three of these features gauge the risk of future trade automation by calculating the profit or loss you would have experienced creating your model in the past and then trading up until the present time.
Out of sample evaluation builds the system on one set of price history, but then evaluates the finished system on a completely different set of prices never used during system development.
Paper trading optimization uses two sets of price history. It optimizes the system on the first set of price history, but automatically keeps the interim result which did best on a second set of price history not used during the optimization process.
Finally, NeuroShell Trader’s walk forward optimization performs multiple optimizations using both out of sample evaluation and paper trading optimization, but it keeps walking each subsequent optimization forward in time to simulate what you would do in real life when re-optimizing with the latest data as market conditions change.
(Image of cross-validation, a key concept of walk-forward algorithms. Image courtesy of Wikipedia)
Q: I noticed that your platform also has the ability for further customization. Are there any languages that your platform supports so that a user could completely code their own trading strategy or alpha signals?
A: Yes, you can program your own indicators and trading systems in a Dynamic Link Library (DLL) using standard programming languages such as C, C++, Power Basic and Delphi. This is the same interface that many developers have used to create 3rd-party add-ons for NeuroShell Trader. However, our users tell us that NeuroShell Trader’s wizards are so powerful that they can easily create custom indicators and trading systems without having to do any coding just by using the wizards in combination with some of the over 800 built-in indicators.
Q: You also offer several different versions of your offering. Can you point out the difference of the ‘Power User’ edition versus the ‘NeuroShell Trader Forex’ edition?
A: There are three different versions of NeuroShell Trader Forex, all of which work exclusively with FXCM.
The basic NeuroShell Trader Forex is $19.99 a month. It includes 250 technical indicators and allows you to easily build neural network predictions and advanced trading systems. It also includes genetic system optimization to help you find the best indicator parameters and trading rules. You can build custom trading models for multiple securities in one pass through the prediction and trading strategy wizards. It also allows you to create alerts and automatically trade your models.
NeuroShell Trader Forex Professional is $49.99 a month. It adds 500 more technical indicators and additional options such as saving of custom indicators and templates, multicore optimization, paper trading optimization, out-of-sample back testing, cross market data analysis, and a few other advanced user options.
NeuroShell Trader Forex Power User is $99.99 a month. It adds multiple time frame analysis, money management, pyramiding and position scaling, walk forward optimization, and batch processing, making it one of the most powerful trading systems on the market.
For a more in depth explanation of the features and differences between the versions, it’s best to visit the NeuroShell Trader Forex website at http://nsforex.wardsystemsgroup.com/
Risk Warning: The FXCM Group does not guarantee accuracy and will not accept liability for any loss or damage which arise directly or indirectly from use of or reliance on information contained within the webinars. The FXCM Group may provide general commentary which is not intended as investment advice and must not be construed as such. FX/CFD trading carries a risk of losses in excess of your deposited funds and may not be suitable for all investors. Please ensure that you fully understand the risks involved.