DOWNLOADS

RTS5Pattern Profi – Advanced Pattern Analysis Indicator for MetaTrader 5

RTS5Pattern is a professional tool for MetaTrader 5 that utilizes statistical matching of historical price formations. With just two clicks – on the first and last candle of your chosen pattern – the indicator automatically scans historical data to find the most similar situations and generates predictions based on their outcomes.

This approach empowers traders to make decisions based on recurring price behavior rather than relying on speculative models or fixed templates. RTSPattern is ideal for price action traders who want to combine visual market analysis with statistically grounded forecasts.

The tool includes predictive visualization, a model of the most statistically favorable outcome, and navigation buttons that allow users to step through detected historical patterns. The indicator is available as a demo version valid until May 30, 2025.

This indicator, RFNN (Remind Fast Neural Network), is based on a neural network architecture, specifically using a Back Propagation Neural Network (BPN). Below is an updated guide on how to use it, based on the provided code.

1. Basic Setup

• Neural Network Inputs: The neural network uses data such as Open, High, Low, Close, along with additional features like DayOfWeek, TimeHour, TimeMinute, and Volume.
• Neural Network Output: The output is computed as the average of the Open, High, Low, and Close values: Output=Open+High+Low+Close4text{Output} = frac{text{Open} + text{High} + text{Low} + text{Close}}{4}Output=4Open+High+Low+Close​
• Directory Setup: For full functionality, the directory C:ANN must be created.

2. Training the Neural Network

• Training Period: The number of bars (candlesticks) used for training is set by BarsTrain (default is 5000).
• Neural Network Layers:
  • The network has two hidden layers with nn_hidden1 and nn_hidden2 neurons, respectively.
  • The total number of input neurons is calculated based on the Inputs parameter, which defines how many bars are used as inputs.
• Starting Training: Pressing the „RUN TRAINING“ button will start the training process. The button’s text changes to „LEARNING IN PROGRESS“ during active training. The network is trained iteratively using past market data to predict future price movements.

3. Future Predictions

• Prediction Period: The network predicts future prices based on past data, and the FuturePredictionShift parameter (default 3) defines how many bars ahead the predictions will be made.
• Prediction Activation: The „BUTON-PREDICT“ button triggers the prediction process, and the predicted values are plotted on the chart.

4. Data Normalization and Denormalization

• Normalization: The input data (Open, High, Low, Close, etc.) is normalized using maximum and minimum values to scale them into the range between 0 and 1, ensuring stable and consistent neural network performance.
• Denormalization: After the network computes the outputs, the predicted values are denormalized to revert them back to the original price scale.

5. Chart Display and Visual Elements

• Several visual elements are plotted on the chart to assist with analyzing the network’s output:
  • NNOutput: Shows the network’s computed output.
  • PredictedBuff: Displays the predicted future prices.
  • DesiredBuff: Represents the target values used during training.
  • NNTrainLineOut: Displays the network’s output during training.
• These elements are represented as arrows or lines, providing visual feedback on the network’s predictions and performance.

6. Saving and Loading Neural Networks

• Saving: The trained neural network can be saved periodically after a set number of iterations.
• Loading: If the Load option is enabled, the last saved network will be loaded from the designated path (C:ANN) upon initialization, allowing for a continuation of the training process without starting from scratch.

7. Indicator Initialization and Timers

• During initialization (init()), the indicator sets up the network parameters, prepares the chart elements (such as buttons and prediction output), and checks the current date for any expiration (in this case, the demo expires on 2024-12-12).
• The training and prediction processes are managed via a timer (OnTimer()), which handles the network’s training loop and ensures that the chart is updated regularly.

8. Manual Adjustments via Chart Events

• Manual Navigation: You can adjust the bar from which the training or prediction starts using the „TI+“ and „TI-“ buttons.
• Training and Prediction Execution: Pressing „BUTON“ starts or stops the training process, while pressing „BUTON-PREDICT“ recalculates the predictions based on the current network.

9. Parallel Processing and Threading

• The indicator supports parallel processing, allowing faster training on multi-threaded systems. This can be useful when dealing with large datasets or when faster computation is required.

10. Practical Notes

• Training Parameters: The neural network’s behavior can be further fine-tuned by adjusting various parameters such as the number of hidden neurons, input layers, or even the activation functions used within the network.
• Visualization: The chart offers immediate feedback on the performance of the network’s predictions, showing both the actual and predicted future prices.

Conclusion:

This indicator provides a sophisticated tool for predicting future price movements using neural networks. Proper setup, training, and tuning of the network can lead to more accurate market predictions, making it a potentially valuable asset for traders.