Mastering the Simple Single Layer Perceptron EA for MetaTrader 4

1. Understanding Perceptrons

Welcome to the world of Perceptrons! If you're curious about how simple neural networks operate, check out this insightful resource by Dr. Mark Humphrys.

In the context of trading, I liken it to the logic behind an Expert Advisor (EA) that leverages these concepts.

2. The Algorithm

Let's break it down.

2.1. Inputs

The variables w1, w2, w3, and w4 represent weights that you’ll set based on your optimization preferences.

input int x1 = 0;//weight1
input int x2 = 0;//weight2
input int x3 = 0;//weight3
input int x4 = 0;//weight4

2.2. The Perceptron

For our simple EA, we set the threshold to zero. The output will be either "fires" (1) or "does not fire" (0).

The Rule

double w1 = x1 - 100;
double w2 = x2 - 100;
double w3 = x3 - 100;
double w4 = x4 - 100;

// Perceptron before one bar 2017/03/18
double a11 = ((iRSI(Symbol(), 0, 12, PRICE_MEDIAN, 1))/100-0.5)*2;
double a21 = ((iRSI(Symbol(), 0, 36, PRICE_MEDIAN, 1))/100-0.5)*2;
double a31 = ((iRSI(Symbol(), 0, 108, PRICE_MEDIAN, 1))/100-0.5)*2;
double a41 = ((iRSI(Symbol(), 0, 324, PRICE_MEDIAN, 1))/100-0.5)*2;
double Current_Percptron = (w1 * a11 + w2 * a21 + w3 * a31 + w4 * a41);

// Perceptron before two bars 2017/03/18
double a12 = ((iRSI(Symbol(), 0, 12, PRICE_MEDIAN, 2))/100-0.5)*2;
double a22 = ((iRSI(Symbol(), 0, 36, PRICE_MEDIAN, 2))/100-0.5)*2;
double a32 = ((iRSI(Symbol(), 0, 108, PRICE_MEDIAN, 2))/100-0.5)*2;
double a42 = ((iRSI(Symbol(), 0, 324, PRICE_MEDIAN, 2))/100-0.5)*2;
double Pre_Percptron = (w1 * a12 + w2 * a22 + w3 * a32 + w4 * a42);

While I've used the RSI indicator for this EA, feel free to experiment with other oscillators like RCI or W%R!

2.3. Order Opening and Closing

If the previous Perceptron is below 0 and the current one is above, close any short positions. This signals the EA to send a long order.

if(Pre_Percptron < 0 && Current_Percptron > 0) // long signal
{
    // Close short position if it exists
    if(pos < 0)
    {
        ret = OrderClose(Ticket, OrderLots(), OrderClosePrice(), 0);
        if(ret) pos = 0; // Position status reset
    }
    // Send long order if no position exists
    if(pos == 0) Ticket = OrderSend(_Symbol, OP_BUY, Lots, Ask, 0, 0, 0, Trade_Comment, MagicNumber, 0, Green);
}

On the flip side, if the current Perceptron is below 0 and the previous one is above, close any long positions and send a short order.

3. Optimization

To optimize, load the "Slime_Mold_RSI_template.set" and select "open price only" for the model.

Inputs

Optimization

4. Comments and Magic Number

The Magic Number is set to the duration used for optimization, and this EA utilizes it in the comments.

string Trade_Comment = IntegerToString(MagicNumber,5,' ') + "Days-Optimization";

Comment

5. Related Article

For further reading, you might find this article helpful: Check it out here!