I think an interesting application of ML could be generating the ‘features’ for inclusion in a trading algorithm, converting non-numerical data into numerical. For example, converting market sentiment or satellite imagery to count cars in a retailer’s parking lot. More examples here. I had written previously about Machine Learning for trading strategy design here.
Thanks to code from @johnromero via @quantocracy I wrote a quick script to draw support and resistance lines automatically using machine learning on any chart. Currently it is horizontal lines but I’m sure trend lines could be done as well.
And if someone decides to create a trading strategy designed around breakouts or mean-reversion, please share results good or bad!