• 概述
  • 评论
  • 评论
  • 新特性

Smart Trend Indicator

Smart Trend Indicator is a trading system that employs the artificial intelligence technologies. The logic and calculations of the indicator are based on mathematical modeling of the trend. To identify the trend, the indicator solves the problem of binary classification in a given range of history data by constructing an adaptive parametric model of the trend. The structure and composition of the main functional elements of the model is similar to that of the neural network model. The optimal trend reversals points of different depth are used as the object of modeling. Predictors are homogeneous data sets consisting of the values of one technical indicator with a fixed time shift.


Customizable Settings

  • History - the number of bars for modeling.
  • InputType - type of input data.
  • InputLags - the number of components in the input data with a fixed time shift.
  • OutputType - type of trend calculation.
  • OutputDepth - depth of the trend.
  • NormalizationType - type of input data normalization.
  • HiddensLayers - the number of hidden layers.
  • Neurons1 - the number of structural elements in the first hidden layer of the model.
  • Neurons2 - the number of structural elements in the second hidden layer of the model.
  • TrainingAlgorithm - training algorithm (0 - LM, 1 - BFGS).
  • BFGSStep - step size.
  • BFGSMaxit - the number of BFGS iterations.
  • TrainingCycles - the number of training cycles.
  • TrainingError - training error.
  • Retrain - retraining mode.
  • RetrainPeriod - the number of bars to perform retraining.
  • SignalLevelOptimization - enable optimization of the signal level.
  • SignalLevel - the minimum signal level.
  • VisualSignalZone - type of signal visualization.
  • ArrowGap - offset of the signal label.
  • ArrowWidth - size (width) of the signal label.
  • InfoBlock - show the information panel on the chart.
无评论
版本 5.5 2018.06.08
1. Added new training algorithm - BFGS;
2. Added the ability for the indicator to work in the retraining mode;
3. Added new training models with extended structure, including up to three hidden layers;
4. Improved visualization of signals.