Making a ML algorithm for trading

"I am curious whether this concept has already been explored and, if not, whether it is feasible and worthwhile to pursue. The idea involves constructing a machine learning algorithm that trains a model on a large dataset of technical chart patterns, such as head and shoulders, flag breakouts, etc., to derive valuable insights. For instance, the algorithm could analyze occurrences and reveal that, say, 75% of the time, a confirmed head and shoulders pattern results in an upward movement of the underlying asset. Additionally, the system could eventually execute trades automatically based on these insights, potentially incorporating dynamic behavior and continuous learning to refine its predictions and adapt to new data over time."