Where in the chart does AI say "buy"? I will show you everything inside.

Many products end by saying “AI will decide.” We do the opposite. We will disclose everything about what the AI of ‘GENSEN-AI’ takes from the chart, and how it leads to a “buy,” in every possible detail.

Input is generated from your MT5 candlesticks. Specifically, it is as follows.

• The most recent 64 candlesticksconverted into a sequence of four series (notation on Netron is 1×64×4)
• Separately, from auxiliary indicators etc.,12 values(also 1×12)

These two are all the inputs to the neural network. Instead of vague descriptions like “analyze all market information,” it is a concrete form of 64×4 sequences and 12 numbers.

We also describe the structure on the receiving end. The 64 sequences pass through three convolution layers (Conv) to compress into features, where 12 numbers are concatenated. After concatenation, they pass through two fully connected layers (Gemm), and finally a Softmax yields three probabilities. These are the “buy,” “sell,” and “hold.” A signal is issued only when the probability exceeds a threshold (initial value 0.40).

What’s important here is that this calculation structure is not hidden. Once training finishes, an ONNX file is generated on your PC, and by opening it with Netron, a free tool, you can visually confirm the layer arrangement, input dimensions, and output shape yourself.

Layer structure opened in Netron. Input 1×64×4 and 1×12, output: Softmax for 3 classes

Of course, even by looking at the structure, we cannot fully explain “why this moment is a buy” in human terms. This is the nature of neural network technology. Therefore, instead of explanations, we provide validation. We determine all signals using about 2,000 candlesticks not used in training, and confirm its performance as a ratio where the number of decisive signals is the denominator, numerically verifying whether this judgment holds in the latest market. Disclosure of contents and machine-compiled performance. If both are in place, there is no need to rely on the belief that “AI decides.”