What does advise() with PATTERN option do in terms of machine learning? Conditions in if() statements in the output code look like made by a rule learner (separate-and-conquer, in contrast to divide-and-conquer by decision tree); values at the leaves look like those made by regression tree. Does Zorro's method have a common name in machine learning world?

I started digging into the topic after trying to compare Karsten Martiny's unsupervised candle pattern discovery with agglomerative clustering (mentioned on robotwealth.com) and http://www.zorro-trader.com/manual/en/advisor.htm

A similar question - does anybody know what PriceActionLab does in terms of machine learning? I'm trying to avoid reverse engineering their long manual http://www.priceactionlab.com/Literature/pal71manual/pal71manual.html

Yes, it is a rule learner. It's not machine learning in the strong sense since it's limited to comparisons. The values at the leaves are determined from the frequency and predictive strength of the rule. Clustering is not used. PriceActionLab is a simplified version of the same method. It just stores the comparison rules that precede simulated profitable trades.

Does this rule learner have a common name in a literature, like VFDR or SLIPPER? Up to now I only tried once the classification rule learner RIPPER (JRip from RWeka R package).

What happens if Objective!=0 (e.g. scaled next bar range, for whatever purpose) is supplied to advise() along with PATTERN? I think it's not a classification problem any more then.

Also for the case Objective==0 and PATTERN, it would be interesting to know how exactly values at leaves are determined from frequency and predictive strength.

It has no particular name, but as you mention it, maybe we should come up with some fancy name for marketing reasons.

Objective = 0 means that the next trade result is the training objective. The values at the leaves are not calculated by a explicit formula, but by some subroutine that, as far as I understand it, sorts the patterns and then retrieves the value from the position in the sort order.