LSTM Neural Network - MetaTrader 5 Library | Forex Indicator Download - MT4/MT5 Resources

The attachment is the include file of LSTM. The files included are:

Also included is a sample LSTMTest script using Simple RPC metrics (also attached).

To create a new LSTM network, provide the constructor with the number of modes, the number of inputs (predictors for each time step), and the number of time steps, as shown below;

 CLSTM network*net= new LSTM network (mode, input, time step);

To train the network, call the Learn function, providing it with the input array, target array, learning error threshold, and number of learning epochs, as shown below;

 net.Learn(in,tg,mse, 500,000 );

After learning is completed, the final error and convergence period can be obtained from the network, as shown below;

 net.MSE();
net.Epochs();

To check for a specific pattern against the network, call the compute function, passing the candidate patterns as arguments to the function, as shown;

 net.calc(in);

The calculation function returns the output. This LSTM has one output neuron.

If anyone finds bugs or has improvements or any suggestions please share. Good luck.

Attachment download

📎 gates.mqh (4.3 KB)

📎 timestep.mqh (3.79 KB)

📎 lstmnetwork.mqh (11.24 KB)

📎 lstmtest.mq5 (3.14 KB)

📎 simple_rpc.mq5 (9.92 KB)

Source: MQL5 #24200