Parameters and Recommended Settings

The Cluster Indicators with Contribution Factors add the capability of letting you decide which of the inputs should have more impact when building the model.  You do this by selecting values for the contribution factors (described below).

There are eight Cluster Indicators with contribution factors. They require from two to nine inputs. These indicators are named as ClusterC2, ClusterC3, …, ClusterC9. The number in the indicator name denotes the number of inputs. The difference between them is in the number of input variables, the number of cluster center dimensions, and the number of contribution factors. There are as many cluster center dimensions and contribution factors as there are inputs. For example, ClusterC5 takes five inputs and finds a cluster center in a five dimensional space with five contribution factors.

Here are all of the parameters for Cluster Indicators with contribution factors along with recommended range settings you should use with the optimizer. The value of N is the number of the Cluster Indicator (e.g. when looking at ClusterC5 then N=5):

Input 1, Input 2, …, Input N – these parameters are the indicator inputs. Our favorites are the indicators in the Price Momentum category, or the Volume category, or Linear Time Regression: Coefficient of Regression (slope) in the Regression category. However, many other indicators should work well too. Let the optimizer find the parameters of your input indicators.

Scale - the look-back moving window size (number of recent bars on which to find the cluster center). We recommend between 20 and 200. The default value is 100.

C1, C2, …, CN – these parameters are the cluster center coordinates which together represent the cluster center. There is one cluster center, but each dimension of the cluster center is represented by an input variable. The genetic optimizer finds these parameters while maximizing the profit. Default values are 0.5, and the range of variation during optimization is from  –1 to 1.

Note:  There is no practical way for you to set this parameter.  That is why Cluster Indicators require either the NeuroShell Trader Professional or the NeuroShell DayTrader Professional because they include the genetic optimizer.

Contrib1, Contrib2, …, ContribN – these parameters are the contribution factors. There is one contribution factor per each input variable. The greater the value, the more important that input will be when measuring "closeness" to the cluster center. The genetic algorithm optimizer finds these parameters while maximizing the profit. Default values are 0.5, and the range of variation during optimization is from 0 to 1.