Classifying market signals (patterns) into trading actions. For instance, a model might output “BUY”, “SELL”, or “HOLD” based on input features. This turns forecasting into a classification problem. Signal classification models are trained on past data labeled with what the “correct” action would have been. Tools like predict_proba can then produce probabilities for each action. Such models are used to generate automated trading signals – e.g. a neural net might classify a candle pattern as bullish or bearish, guiding the expert advisor.

Back to Robots Glossary Index