A neural network is a series of algorithms that attempts to identify underlying relationships in a set of data by using a process that mimics the way the human brain operates. Neural networks have the ability to adapt to changing input so the network produces the best possible result without the need to redesign the output criteria. The concept of neural networks is rapidly increasing in popularity in the area of developing trading systems.
A neural network operates similar to the brain’s neural network. A “neuron” in a neural network is a simple mathematical function capturing and organizing information according to an architecture. The network closely resembles statistical methods such as curve fitting and regression analysis.
A neural network consists of layers of interconnected nodes. Each node is a perceptron and resembles a multiple linear regression. The perceptron feeds the signal generated by a multiple linear regression into an activation function that may be nonlinear.
In a multi-layered perceptron (MLP), perceptrons are arranged in interconnected layers. The input layer receives input patterns. The output layer contains classifications or output signals to which input patterns may map. For example, the patterns may be a list of quantities for technical indicators regarding a security; potential outputs could be “buy,” “hold” or “sell.” Hidden layers adjust the weightings on the inputs until the error of the neural network is minimal. It is theorized that hidden layers extract salient features in the input data that have predictive power with respect to the outputs. This describes feature extraction, which performs a function similar to statistical techniques such as principal component analysis.
Neural networks are widely used in financial operations, enterprise planning, trading, business analytics and product maintenance. Neural networks are common in business applications such as forecasting and marketing research solutions, fraud detection and risk assessment.
A neural network analyzes price data and uncovers opportunities for making trade decisions based on thoroughly analyzed data. The networks can detect subtle nonlinear interdependencies and patterns other methods of technical analysis cannot uncover. However, a 10% increase in efficiency is all an investor can expect from a neural network. There will always be data sets and task classes for which previously used algorithms remain superior. The algorithm is not what matters; it is the well-prepared input information on the targeted indicator that determines the success of a neural network.